spb/doc-api

1---2title: Sessions3url: https://platform.claude.com/docs/en/api/beta/sessions4---56# Sessions78## Create Session910**POST** `/v1/sessions`1112Create Session1314### Headers1516- `"anthropic-beta": optional array of AnthropicBeta`1718  Optional header to specify the beta version(s) you want to use.1920  - `string`2122  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`2324    - `"message-batches-2024-09-24"`2526    - `"prompt-caching-2024-07-31"`2728    - `"computer-use-2024-10-22"`2930    - `"computer-use-2025-01-24"`3132    - `"pdfs-2024-09-25"`3334    - `"token-counting-2024-11-01"`3536    - `"token-efficient-tools-2025-02-19"`3738    - `"output-128k-2025-02-19"`3940    - `"files-api-2025-04-14"`4142    - `"mcp-client-2025-04-04"`4344    - `"mcp-client-2025-11-20"`4546    - `"dev-full-thinking-2025-05-14"`4748    - `"interleaved-thinking-2025-05-14"`4950    - `"code-execution-2025-05-22"`5152    - `"extended-cache-ttl-2025-04-11"`5354    - `"context-1m-2025-08-07"`5556    - `"context-management-2025-06-27"`5758    - `"model-context-window-exceeded-2025-08-26"`5960    - `"skills-2025-10-02"`6162    - `"fast-mode-2026-02-01"`6364    - `"output-300k-2026-03-24"`6566    - `"user-profiles-2026-03-24"`6768    - `"user-profiles-2026-08-18"`6970    - `"user-profiles-2026-09-04"`7172    - `"advisor-tool-2026-03-01"`7374    - `"managed-agents-2026-04-01"`7576    - `"cache-diagnosis-2026-04-07"`7778    - `"dreaming-2026-04-21"`7980    - `"thinking-token-count-2026-05-13"`8182    - `"server-side-fallback-2026-06-01"`8384    - `"server-side-fallback-2026-07-01"`8586    - `"fallback-credit-2026-06-01"`8788    - `"fallback-credit-2026-07-01"`8990    - `"agent-memory-2026-07-22"`9192    - `"mid-conversation-tool-changes-2026-07-01"`9394    - `"compact-2026-01-12"`9596    - `"computer-use-2025-11-24"`9798    - `"mcp-tunnels-2026-06-22"`99100    - `"structured-outputs-2025-11-13"`101102    - `"task-budgets-2026-03-13"`103104    - `"thinking-display-updates-2026-08-18"`105106    - `"ce-user-management-2026-07-13"`107108    - `"mid-conversation-output-config-2026-07-01"`109110    - `"thinking-binding-controls-2026-08-01"`111112    - `"mid-conversation-system-clear-at-2026-08-21"`113114    - `"compact-2026-09-04"`115116- `"anthropic-workspace-id": optional string`117118### Body parameters119120- `agent: string or BetaManagedAgentsAgentParams or BetaManagedAgentsAgentWithOverridesParams`121122  Agent identifier. Accepts the `agent` ID string, which pins the latest version for the session, or an `agent` object with both id and version specified.123124  - `string`125126  - `BetaManagedAgentsAgentParams object`127128    Specification for an Agent. Provide a specific `version` or use the short-form `agent="agent_id"` for the most recent version129130    - `type: "agent"`131132    - `id: string`133134      The `agent` ID.135136      minLength: 1, maxLength: 128137138    - `version: optional number`139140      The specific `agent` version to use. Omit to use the latest version. Must be at least 1 if specified.141142      format: int32143144  - `BetaManagedAgentsAgentWithOverridesParams object`145146    Reference to an `agent` plus optional configuration overrides. Each provided field replaces the agent's value for the caller's use; the agent resource is unchanged.147148    - `type: "agent_with_overrides"`149150    - `id: string`151152      The `agent` ID.153154      minLength: 1, maxLength: 128155156    - `mcp_servers: optional array of BetaManagedAgentsURLMCPServerParams`157158      Replacement MCP server list. Full replacement: the provided array becomes the MCP servers. Send an empty array to clear; omit to preserve the agent's servers.159160      - `type: "url"`161162      - `name: string`163164        Unique name for this server, referenced by mcp_toolset configurations. 1-255 characters.165166        minLength: 1, maxLength: 255167168      - `url: string`169170        Endpoint URL for the MCP server.171172        maxLength: 2048173174    - `model: optional BetaManagedAgentsModel or BetaManagedAgentsModelConfigParams`175176      Replacement model. Accepts the model string, e.g. `claude-opus-5`, or a `model_config` object. Omit to use the agent's model.177178      - `BetaManagedAgentsModel = "claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more or string`179180        The model that will power your agent.181182        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.183184        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`185186          The model that will power your agent.187188          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.189190          - `"claude-fable-5-1"`191192            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows193194          - `"claude-sonnet-5"`195196            High-performance model for coding and agents197198          - `"claude-fable-5"`199200            Next generation of intelligence for the hardest knowledge work and coding problems201202          - `"claude-opus-5"`203204            Powerful intelligence for long-running agents and coding205206          - `"claude-opus-4-8"`207208            Powerful intelligence for long-running agents and coding209210          - `"claude-opus-4-7"`211212            Powerful intelligence for long-running agents and coding213214          - `"claude-opus-4-6"`215216            Powerful intelligence for long-running agents and coding217218          - `"claude-sonnet-4-6"`219220            Best combination of speed and intelligence221222          - `"claude-haiku-4-5"`223224            Fastest model with near-frontier intelligence225226          - `"claude-haiku-4-5-20251001"`227228            Fastest model with near-frontier intelligence229230          - `"claude-opus-4-5"`231232            Powerful intelligence for long-running agents and coding233234          - `"claude-opus-4-5-20251101"`235236            Powerful intelligence for long-running agents and coding237238          - `"claude-sonnet-4-5"`239240            High-performance model for agents and coding241242          - `"claude-sonnet-4-5-20250929"`243244            High-performance model for agents and coding245246        - `string`247248      - `BetaManagedAgentsModelConfigParams object`249250        An object that defines additional configuration control over model use251252        - `id: BetaManagedAgentsModel`253254          The model that will power your agent.255256          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.257258        - `effort: optional "low" or "medium" or "high" or 2 more or BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or 3 more or null`259260          How hard Claude works on each inference call. Accepts a bare level string (`"high"`) or `{"type": "high"}`. On create, omitting it resolves the per-model default; on update, omitting it leaves the stored value unchanged.261262          - `BetaManagedAgentsEffortLevel = "low" or "medium" or "high" or 2 more`263264            How hard Claude works on each turn. Higher levels favor reasoning depth over latency. Not all models accept every level; invalid combinations are rejected at create time.265266            - `"low"`267268            - `"medium"`269270            - `"high"`271272            - `"xhigh"`273274            - `"max"`275276          - `BetaManagedAgentsEffortLow object`277278            Low effort. Favors latency over reasoning depth.279280            - `type: "low"`281282          - `BetaManagedAgentsEffortMedium object`283284            Medium effort. Balances latency and reasoning depth.285286            - `type: "medium"`287288          - `BetaManagedAgentsEffortHigh object`289290            High effort. Favors reasoning depth.291292            - `type: "high"`293294          - `BetaManagedAgentsEffortXhigh object`295296            Extra-high effort. Not all models accept this level.297298            - `type: "xhigh"`299300          - `BetaManagedAgentsEffortMax object`301302            Maximum effort. Favors reasoning depth over latency.303304            - `type: "max"`305306        - `inference_geo: optional string or null`307308          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo. On update, `model` is whole-object replacement — omitting inference_geo clears it.309310        - `speed: optional "standard" or "fast" or null`311312          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.313314          - `"standard"`315316          - `"fast"`317318    - `skills: optional array of BetaManagedAgentsSkillParams`319320      Replacement skill list. Full replacement: the provided array becomes the skills. Send an empty array to clear; omit to preserve the agent's skills.321322      - `BetaManagedAgentsAnthropicSkillParams object`323324        An Anthropic-managed skill.325326        - `type: "anthropic"`327328        - `skill_id: string`329330          Identifier of the Anthropic skill (e.g., "xlsx").331332          minLength: 1, maxLength: 64333334        - `version: optional string or null`335336          Version to pin. Defaults to latest if omitted.337338          minLength: 1, maxLength: 64339340      - `BetaManagedAgentsCustomSkillParams object`341342        A user-created custom skill.343344        - `type: "custom"`345346        - `skill_id: string`347348          Tagged ID of the custom skill (e.g., "skill_01XJ5...").349350          minLength: 1, maxLength: 64351352        - `version: optional string or null`353354          Version to pin. Defaults to latest if omitted.355356          minLength: 1, maxLength: 64357358    - `system: optional string or null`359360      Replacement system prompt. Up to 100,000 characters. Set to null to clear the agent's system prompt; omit to preserve it.361362      maxLength: 100000363364    - `tools: optional array of BetaManagedAgentsAgentToolset20260401Params or BetaManagedAgentsMCPToolsetParams or BetaManagedAgentsCustomToolParams`365366      Replacement tool list. Full replacement: the provided array becomes the tool configuration. Send an empty array to clear; omit to preserve the agent's tools.367368      - `BetaManagedAgentsAgentToolset20260401Params object`369370        Configuration for built-in agent tools. Use this to enable or disable groups of tools available to the agent.371372        - `type: "agent_toolset_20260401"`373374        - `configs: optional array of BetaManagedAgentsAgentToolConfigParams`375376          Per-tool configuration overrides.377378          - `BetaManagedAgentsBashToolConfigParams object`379380            Configuration override for the bash tool.381382            - `type: optional "bash"`383384            - `name: "bash"`385386              Must be "bash".387388            - `enabled: optional boolean or null`389390              Whether this tool is enabled and available to Claude. Overrides the default_config setting.391392            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`393394              Permission policy for tool execution.395396              - `BetaManagedAgentsAlwaysAllowPolicy object`397398                Tool calls are automatically approved without user confirmation.399400                - `type: "always_allow"`401402              - `BetaManagedAgentsAlwaysAskPolicy object`403404                Tool calls require user confirmation before execution.405406                - `type: "always_ask"`407408              - `BetaManagedAgentsAutoPolicy object`409410                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.411412                - `type: "auto"`413414          - `BetaManagedAgentsEditToolConfigParams object`415416            Configuration override for the edit tool.417418            - `type: optional "edit"`419420            - `name: "edit"`421422              Must be "edit".423424            - `enabled: optional boolean or null`425426              Whether this tool is enabled and available to Claude. Overrides the default_config setting.427428            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`429430              Permission policy for tool execution.431432              - `BetaManagedAgentsAlwaysAllowPolicy object`433434                Tool calls are automatically approved without user confirmation.435436              - `BetaManagedAgentsAlwaysAskPolicy object`437438                Tool calls require user confirmation before execution.439440              - `BetaManagedAgentsAutoPolicy object`441442                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.443444          - `BetaManagedAgentsReadToolConfigParams object`445446            Configuration override for the read tool.447448            - `type: optional "read"`449450            - `name: "read"`451452              Must be "read".453454            - `enabled: optional boolean or null`455456              Whether this tool is enabled and available to Claude. Overrides the default_config setting.457458            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`459460              Permission policy for tool execution.461462              - `BetaManagedAgentsAlwaysAllowPolicy object`463464                Tool calls are automatically approved without user confirmation.465466              - `BetaManagedAgentsAlwaysAskPolicy object`467468                Tool calls require user confirmation before execution.469470              - `BetaManagedAgentsAutoPolicy object`471472                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.473474          - `BetaManagedAgentsWriteToolConfigParams object`475476            Configuration override for the write tool.477478            - `type: optional "write"`479480            - `name: "write"`481482              Must be "write".483484            - `enabled: optional boolean or null`485486              Whether this tool is enabled and available to Claude. Overrides the default_config setting.487488            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`489490              Permission policy for tool execution.491492              - `BetaManagedAgentsAlwaysAllowPolicy object`493494                Tool calls are automatically approved without user confirmation.495496              - `BetaManagedAgentsAlwaysAskPolicy object`497498                Tool calls require user confirmation before execution.499500              - `BetaManagedAgentsAutoPolicy object`501502                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.503504          - `BetaManagedAgentsGlobToolConfigParams object`505506            Configuration override for the glob tool.507508            - `type: optional "glob"`509510            - `name: "glob"`511512              Must be "glob".513514            - `enabled: optional boolean or null`515516              Whether this tool is enabled and available to Claude. Overrides the default_config setting.517518            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`519520              Permission policy for tool execution.521522              - `BetaManagedAgentsAlwaysAllowPolicy object`523524                Tool calls are automatically approved without user confirmation.525526              - `BetaManagedAgentsAlwaysAskPolicy object`527528                Tool calls require user confirmation before execution.529530              - `BetaManagedAgentsAutoPolicy object`531532                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.533534          - `BetaManagedAgentsGrepToolConfigParams object`535536            Configuration override for the grep tool.537538            - `type: optional "grep"`539540            - `name: "grep"`541542              Must be "grep".543544            - `enabled: optional boolean or null`545546              Whether this tool is enabled and available to Claude. Overrides the default_config setting.547548            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`549550              Permission policy for tool execution.551552              - `BetaManagedAgentsAlwaysAllowPolicy object`553554                Tool calls are automatically approved without user confirmation.555556              - `BetaManagedAgentsAlwaysAskPolicy object`557558                Tool calls require user confirmation before execution.559560              - `BetaManagedAgentsAutoPolicy object`561562                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.563564          - `BetaManagedAgentsWebFetchToolConfigParams object`565566            Configuration override for the web_fetch tool.567568            - `type: optional "web_fetch"`569570            - `name: "web_fetch"`571572              Must be "web_fetch".573574            - `allowed_domains: optional array of string`575576              Only fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.577578            - `blocked_domains: optional array of string`579580              Never fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.581582            - `enabled: optional boolean or null`583584              Whether this tool is enabled and available to Claude. Overrides the default_config setting.585586            - `max_content_tokens: optional number or null`587588              Maximum number of tokens of fetched text content to include in context per call. Does not apply to binary content such as PDFs.589590              format: int32591592            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`593594              Permission policy for tool execution.595596              - `BetaManagedAgentsAlwaysAllowPolicy object`597598                Tool calls are automatically approved without user confirmation.599600              - `BetaManagedAgentsAlwaysAskPolicy object`601602                Tool calls require user confirmation before execution.603604              - `BetaManagedAgentsAutoPolicy object`605606                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.607608          - `BetaManagedAgentsWebSearchToolConfigParams object`609610            Configuration override for the web_search tool.611612            - `type: optional "web_search"`613614            - `name: "web_search"`615616              Must be "web_search".617618            - `allowed_domains: optional array of string`619620              Only return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.621622            - `blocked_domains: optional array of string`623624              Never return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.625626            - `enabled: optional boolean or null`627628              Whether this tool is enabled and available to Claude. Overrides the default_config setting.629630            - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`631632              Permission policy for tool execution.633634              - `BetaManagedAgentsAlwaysAllowPolicy object`635636                Tool calls are automatically approved without user confirmation.637638              - `BetaManagedAgentsAlwaysAskPolicy object`639640                Tool calls require user confirmation before execution.641642              - `BetaManagedAgentsAutoPolicy object`643644                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.645646            - `user_location: optional BetaManagedAgentsUserLocation or null`647648              Approximate user location for search result localization.649650              - `type: "approximate"`651652                Location precision. Only "approximate" is supported.653654              - `city: optional string or null`655656                City name.657658                minLength: 1, maxLength: 255659660              - `country: optional string or null`661662                Two-letter ISO 3166-1 country code, uppercase.663664              - `region: optional string or null`665666                Region or state name.667668                minLength: 1, maxLength: 255669670              - `timezone: optional string or null`671672                IANA timezone identifier, e.g. "America/Los_Angeles".673674                minLength: 1, maxLength: 255675676        - `default_config: optional BetaManagedAgentsAgentToolsetDefaultConfigParams or null`677678          Default configuration for all tools in a toolset.679680          - `enabled: optional boolean or null`681682            Whether tools are enabled and available to Claude by default. Defaults to true if not specified.683684          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`685686            Permission policy for tool execution.687688            - `BetaManagedAgentsAlwaysAllowPolicy object`689690              Tool calls are automatically approved without user confirmation.691692            - `BetaManagedAgentsAlwaysAskPolicy object`693694              Tool calls require user confirmation before execution.695696            - `BetaManagedAgentsAutoPolicy object`697698              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.699700      - `BetaManagedAgentsMCPToolsetParams object`701702        Configuration for tools from an MCP server defined in `mcp_servers`.703704        - `type: "mcp_toolset"`705706        - `mcp_server_name: string`707708          Name of the MCP server. Must match a server name from the mcp_servers array. 1-255 characters.709710          minLength: 1, maxLength: 255711712        - `configs: optional array of BetaManagedAgentsMCPToolConfigParams`713714          Per-tool configuration overrides.715716          - `name: string`717718            Name of the MCP tool to configure. 1-128 characters.719720            minLength: 1, maxLength: 128721722          - `enabled: optional boolean or null`723724            Whether this tool is enabled. Overrides the `default_config` setting.725726          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`727728            Permission policy for tool execution.729730            - `BetaManagedAgentsAlwaysAllowPolicy object`731732              Tool calls are automatically approved without user confirmation.733734            - `BetaManagedAgentsAlwaysAskPolicy object`735736              Tool calls require user confirmation before execution.737738            - `BetaManagedAgentsAutoPolicy object`739740              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.741742        - `default_config: optional BetaManagedAgentsMCPToolsetDefaultConfigParams or null`743744          Default configuration for all tools from an MCP server.745746          - `enabled: optional boolean or null`747748            Whether tools are enabled by default. Defaults to true if not specified.749750          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`751752            Permission policy for tool execution.753754            - `BetaManagedAgentsAlwaysAllowPolicy object`755756              Tool calls are automatically approved without user confirmation.757758            - `BetaManagedAgentsAlwaysAskPolicy object`759760              Tool calls require user confirmation before execution.761762            - `BetaManagedAgentsAutoPolicy object`763764              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.765766      - `BetaManagedAgentsCustomToolParams object`767768        A custom tool that is executed by the API client rather than the agent. When the agent calls this tool, an `agent.custom_tool_use` event is emitted and the session goes idle, waiting for the client to provide the result via a `user.custom_tool_result` event.769770        - `type: "custom"`771772        - `description: string`773774          Description of what the tool does, shown to the agent to help it decide when to use the tool.775776          minLength: 1777778        - `input_schema: BetaManagedAgentsCustomToolInputSchema`779780          JSON Schema for custom tool input parameters.781782          - `type: "object"`783784          - `properties: optional map[unknown] or null`785786          - `required: optional array of string or null`787788        - `name: string`789790          Unique name for the tool. 1-128 characters; letters, digits, underscores, and hyphens.791792          minLength: 1, maxLength: 128793794    - `version: optional number`795796      The specific `agent` version to use. Omit to use the latest version.797798      format: int32799800- `environment_id: string`801802  ID of the `environment` defining the container configuration for this session.803804  minLength: 1, maxLength: 128805806- `budget: optional BetaManagedAgentsBudgetLimit`807808  A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.809810  - `type: "limit"`811812  - `max_list_cost: BetaMonetaryAmount`813814    A monetary amount in a specific currency.815816    - `amount: string`817818      Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.819820    - `currency: BetaCurrency`821822      Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.823824- `initial_events: optional array of BetaManagedAgentsUserMessageEventParams or BetaManagedAgentsUserDefineOutcomeEventParams`825826  Initial events to send to the `session` at creation, processed in order. Supports `user.message` and `user.define_outcome` events. Maximum 50 events.827828  - `BetaManagedAgentsUserMessageEventParams object`829830    Parameters for sending a user message to the session.831832    - `type: "user.message"`833834    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`835836      Array of content blocks for the user message.837838      - `BetaManagedAgentsTextBlock object`839840        Regular text content.841842        - `type: "text"`843844        - `text: string`845846          The text content.847848          minLength: 1849850      - `BetaManagedAgentsImageBlock object`851852        Image content specified directly as base64 data or as a reference via a URL.853854        - `type: "image"`855856        - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`857858          Union type for image source variants.859860          - `BetaManagedAgentsBase64ImageSource object`861862            Base64-encoded image data.863864            - `type: "base64"`865866            - `data: string`867868              Base64-encoded image data.869870              minLength: 1871872            - `media_type: string`873874              MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").875876              minLength: 1877878          - `BetaManagedAgentsURLImageSource object`879880            Image referenced by URL.881882            - `type: "url"`883884            - `url: string`885886              URL of the image to fetch.887888              minLength: 1889890          - `BetaManagedAgentsFileImageSource object`891892            Image referenced by file ID.893894            - `type: "file"`895896            - `file_id: string`897898              ID of a previously uploaded file.899900              minLength: 1901902      - `BetaManagedAgentsDocumentBlock object`903904        Document content, either specified directly as base64 data, as text, or as a reference via a URL.905906        - `type: "document"`907908        - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`909910          Union type for document source variants.911912          - `BetaManagedAgentsBase64DocumentSource object`913914            Base64-encoded document data.915916            - `type: "base64"`917918            - `data: string`919920              Base64-encoded document data.921922              minLength: 1923924            - `media_type: string`925926              MIME type of the document (e.g., "application/pdf").927928              minLength: 1929930          - `BetaManagedAgentsPlainTextDocumentSource object`931932            Plain text document content.933934            - `type: "text"`935936            - `data: string`937938              The plain text content.939940              minLength: 1941942            - `media_type: "text/plain"`943944              MIME type of the text content. Must be "text/plain".945946          - `BetaManagedAgentsURLDocumentSource object`947948            Document referenced by URL.949950            - `type: "url"`951952            - `url: string`953954              URL of the document to fetch.955956              minLength: 1957958          - `BetaManagedAgentsFileDocumentSource object`959960            Document referenced by file ID.961962            - `type: "file"`963964            - `file_id: string`965966              ID of a previously uploaded file.967968              minLength: 1969970        - `context: optional string or null`971972          Additional context about the document for the model.973974        - `title: optional string or null`975976          The title of the document.977978      - `BetaManagedAgentsRedactedBlock object`979980        Placeholder for content withheld by Anthropic model policy.981982        - `type: "redacted"`983984  - `BetaManagedAgentsUserDefineOutcomeEventParams object`985986    Parameters for defining an outcome the agent should work toward. The agent begins work on receipt.987988    - `type: "user.define_outcome"`989990    - `description: string`991992      What the agent should produce. This is the task specification.993994    - `rubric: BetaManagedAgentsFileRubricParams or BetaManagedAgentsTextRubricParams`995996      Rubric for grading the quality of an outcome.997998      - `BetaManagedAgentsFileRubricParams object`9991000        Rubric referenced by a file uploaded via the Files API.10011002        - `type: "file"`10031004        - `file_id: string`10051006          ID of the rubric file.10071008      - `BetaManagedAgentsTextRubricParams object`10091010        Rubric content provided inline as text.10111012        - `type: "text"`10131014        - `content: string`10151016          Rubric content. Plain text or markdown — the grader treats it as freeform text. Maximum 262144 characters.10171018          maxLength: 26214410191020    - `max_iterations: optional number or null`10211022      Eval→revision cycles before giving up. Default 3, max 20.10231024      format: int3210251026- `metadata: optional map[string]`10271028  Arbitrary key-value metadata attached to the session. Maximum 16 pairs, keys up to 64 chars, values up to 512 chars.10291030- `resources: optional array of BetaManagedAgentsGitHubRepositoryResourceParams or BetaManagedAgentsFileResourceParams or BetaManagedAgentsMemoryStoreResourceParam`10311032  Resources (e.g. repositories, files) to mount into the session's container.10331034  - `BetaManagedAgentsGitHubRepositoryResourceParams object`10351036    Mount a GitHub repository into the session's container.10371038    - `type: "github_repository"`10391040    - `url: string`10411042      Github URL of the repository10431044      minLength: 1, maxLength: 204810451046    - `authorization_token: optional string`10471048      GitHub authorization token used to clone the repository. Required for private repositories; optional for public ones.10491050      minLength: 1, maxLength: 409610511052    - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`10531054      Branch or commit to check out. Defaults to the repository's default branch.10551056      - `BetaManagedAgentsBranchCheckout object`10571058        - `type: "branch"`10591060        - `name: string`10611062          Branch name to check out.10631064          minLength: 1, maxLength: 25510651066      - `BetaManagedAgentsCommitCheckout object`10671068        - `type: "commit"`10691070        - `sha: string`10711072          Full commit SHA to check out.10731074          minLength: 7, maxLength: 6410751076    - `mount_path: optional string or null`10771078      Mount path in the container. Defaults to `/workspace/<repo-name>`.10791080      minLength: 1, maxLength: 409610811082  - `BetaManagedAgentsFileResourceParams object`10831084    Mount a file uploaded via the Files API into the session.10851086    - `type: "file"`10871088    - `file_id: string`10891090      ID of a previously uploaded file.10911092      minLength: 1, maxLength: 12810931094    - `mount_path: optional string or null`10951096      Mount path in the container. Defaults to `/mnt/session/uploads/<file_id>`.10971098      minLength: 1, maxLength: 409610991100  - `BetaManagedAgentsMemoryStoreResourceParam object`11011102    Parameters for attaching a memory store to an agent session.11031104    - `type: "memory_store"`11051106    - `memory_store_id: string`11071108      The memory store ID (memstore_...). Must belong to the caller's organization and workspace.11091110    - `access: optional "read_write" or "read_only" or null`11111112      Access mode for an attached memory store.11131114      - `"read_write"`11151116      - `"read_only"`11171118    - `instructions: optional string or null`11191120      Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.11211122      maxLength: 409611231124- `title: optional string or null`11251126  Human-readable session title.11271128  maxLength: 50011291130- `vault_ids: optional array of string`11311132  Vault IDs for stored credentials the agent can use during the session.11331134### Returns11351136- `BetaManagedAgentsSession object`11371138  A Managed Agents `session`.11391140  - `type: "session"`11411142  - `id: string`11431144  - `agent: BetaManagedAgentsSessionAgent`11451146    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.11471148    - `type: "agent"`11491150    - `id: string`11511152    - `description: string or null`11531154    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`11551156      - `type: "url"`11571158      - `name: string`11591160      - `url: string`11611162    - `model: BetaManagedAgentsModelConfig`11631164      Model identifier and configuration.11651166      - `id: BetaManagedAgentsModel`11671168        The model that will power your agent.11691170        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.11711172        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`11731174          The model that will power your agent.11751176          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.11771178          - `"claude-fable-5-1"`11791180            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows11811182          - `"claude-sonnet-5"`11831184            High-performance model for coding and agents11851186          - `"claude-fable-5"`11871188            Next generation of intelligence for the hardest knowledge work and coding problems11891190          - `"claude-opus-5"`11911192            Powerful intelligence for long-running agents and coding11931194          - `"claude-opus-4-8"`11951196            Powerful intelligence for long-running agents and coding11971198          - `"claude-opus-4-7"`11991200            Powerful intelligence for long-running agents and coding12011202          - `"claude-opus-4-6"`12031204            Powerful intelligence for long-running agents and coding12051206          - `"claude-sonnet-4-6"`12071208            Best combination of speed and intelligence12091210          - `"claude-haiku-4-5"`12111212            Fastest model with near-frontier intelligence12131214          - `"claude-haiku-4-5-20251001"`12151216            Fastest model with near-frontier intelligence12171218          - `"claude-opus-4-5"`12191220            Powerful intelligence for long-running agents and coding12211222          - `"claude-opus-4-5-20251101"`12231224            Powerful intelligence for long-running agents and coding12251226          - `"claude-sonnet-4-5"`12271228            High-performance model for agents and coding12291230          - `"claude-sonnet-4-5-20250929"`12311232            High-performance model for agents and coding12331234        - `string`12351236      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`12371238        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.12391240        - `BetaManagedAgentsEffortLow object`12411242          Low effort. Favors latency over reasoning depth.12431244          - `type: "low"`12451246        - `BetaManagedAgentsEffortMedium object`12471248          Medium effort. Balances latency and reasoning depth.12491250          - `type: "medium"`12511252        - `BetaManagedAgentsEffortHigh object`12531254          High effort. Favors reasoning depth.12551256          - `type: "high"`12571258        - `BetaManagedAgentsEffortXhigh object`12591260          Extra-high effort. Not all models accept this level.12611262          - `type: "xhigh"`12631264        - `BetaManagedAgentsEffortMax object`12651266          Maximum effort. Favors reasoning depth over latency.12671268          - `type: "max"`12691270      - `inference_geo: optional string`12711272        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.12731274      - `speed: optional "standard" or "fast"`12751276        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.12771278        - `"standard"`12791280        - `"fast"`12811282    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`12831284      Resolved coordinator topology with full agent definitions for each roster member.12851286      - `type: "coordinator"`12871288      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`12891290        Full `agent` definitions the coordinator may spawn as session threads.12911292        - `BetaManagedAgentsSessionThreadAgent object`12931294          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.12951296          - `type: "agent"`12971298          - `id: string`12991300          - `description: string or null`13011302          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`13031304            - `type: "url"`13051306            - `name: string`13071308            - `url: string`13091310          - `model: BetaManagedAgentsModelConfig`13111312            Model identifier and configuration.13131314          - `name: string`13151316          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`13171318            - `BetaManagedAgentsAnthropicSkill object`13191320              A resolved Anthropic-managed skill.13211322              - `type: "anthropic"`13231324              - `skill_id: string`13251326              - `version: string`13271328            - `BetaManagedAgentsCustomSkill object`13291330              A resolved user-created custom skill.13311332              - `type: "custom"`13331334              - `skill_id: string`13351336              - `version: string`13371338          - `system: string or null`13391340          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`13411342            - `BetaManagedAgentsAgentToolset20260401 object`13431344              - `type: "agent_toolset_20260401"`13451346              - `configs: array of BetaManagedAgentsAgentToolConfig`13471348                - `BetaManagedAgentsBashToolConfig object`13491350                  Configuration for the bash tool.13511352                  - `type: "bash"`13531354                  - `enabled: boolean`13551356                  - `name: "bash"`13571358                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`13591360                    Permission policy for tool execution.13611362                    - `BetaManagedAgentsAlwaysAllowPolicy object`13631364                      Tool calls are automatically approved without user confirmation.13651366                      - `type: "always_allow"`13671368                    - `BetaManagedAgentsAlwaysAskPolicy object`13691370                      Tool calls require user confirmation before execution.13711372                      - `type: "always_ask"`13731374                    - `BetaManagedAgentsAutoPolicy object`13751376                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.13771378                      - `type: "auto"`13791380                - `BetaManagedAgentsEditToolConfig object`13811382                  Configuration for the edit tool.13831384                  - `type: "edit"`13851386                  - `enabled: boolean`13871388                  - `name: "edit"`13891390                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`13911392                    Permission policy for tool execution.13931394                    - `BetaManagedAgentsAlwaysAllowPolicy object`13951396                      Tool calls are automatically approved without user confirmation.13971398                    - `BetaManagedAgentsAlwaysAskPolicy object`13991400                      Tool calls require user confirmation before execution.14011402                    - `BetaManagedAgentsAutoPolicy object`14031404                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.14051406                - `BetaManagedAgentsReadToolConfig object`14071408                  Configuration for the read tool.14091410                  - `type: "read"`14111412                  - `enabled: boolean`14131414                  - `name: "read"`14151416                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14171418                    Permission policy for tool execution.14191420                    - `BetaManagedAgentsAlwaysAllowPolicy object`14211422                      Tool calls are automatically approved without user confirmation.14231424                    - `BetaManagedAgentsAlwaysAskPolicy object`14251426                      Tool calls require user confirmation before execution.14271428                    - `BetaManagedAgentsAutoPolicy object`14291430                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.14311432                - `BetaManagedAgentsWriteToolConfig object`14331434                  Configuration for the write tool.14351436                  - `type: "write"`14371438                  - `enabled: boolean`14391440                  - `name: "write"`14411442                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14431444                    Permission policy for tool execution.14451446                    - `BetaManagedAgentsAlwaysAllowPolicy object`14471448                      Tool calls are automatically approved without user confirmation.14491450                    - `BetaManagedAgentsAlwaysAskPolicy object`14511452                      Tool calls require user confirmation before execution.14531454                    - `BetaManagedAgentsAutoPolicy object`14551456                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.14571458                - `BetaManagedAgentsGlobToolConfig object`14591460                  Configuration for the glob tool.14611462                  - `type: "glob"`14631464                  - `enabled: boolean`14651466                  - `name: "glob"`14671468                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14691470                    Permission policy for tool execution.14711472                    - `BetaManagedAgentsAlwaysAllowPolicy object`14731474                      Tool calls are automatically approved without user confirmation.14751476                    - `BetaManagedAgentsAlwaysAskPolicy object`14771478                      Tool calls require user confirmation before execution.14791480                    - `BetaManagedAgentsAutoPolicy object`14811482                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.14831484                - `BetaManagedAgentsGrepToolConfig object`14851486                  Configuration for the grep tool.14871488                  - `type: "grep"`14891490                  - `enabled: boolean`14911492                  - `name: "grep"`14931494                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14951496                    Permission policy for tool execution.14971498                    - `BetaManagedAgentsAlwaysAllowPolicy object`14991500                      Tool calls are automatically approved without user confirmation.15011502                    - `BetaManagedAgentsAlwaysAskPolicy object`15031504                      Tool calls require user confirmation before execution.15051506                    - `BetaManagedAgentsAutoPolicy object`15071508                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.15091510                - `BetaManagedAgentsWebFetchToolConfig object`15111512                  Configuration for the web_fetch tool.15131514                  - `type: "web_fetch"`15151516                  - `enabled: boolean`15171518                  - `name: "web_fetch"`15191520                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`15211522                    Permission policy for tool execution.15231524                    - `BetaManagedAgentsAlwaysAllowPolicy object`15251526                      Tool calls are automatically approved without user confirmation.15271528                    - `BetaManagedAgentsAlwaysAskPolicy object`15291530                      Tool calls require user confirmation before execution.15311532                    - `BetaManagedAgentsAutoPolicy object`15331534                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.15351536                  - `allowed_domains: optional array of string`15371538                  - `blocked_domains: optional array of string`15391540                  - `max_content_tokens: optional number or null`15411542                    format: int3215431544                - `BetaManagedAgentsWebSearchToolConfig object`15451546                  Configuration for the web_search tool.15471548                  - `type: "web_search"`15491550                  - `enabled: boolean`15511552                  - `name: "web_search"`15531554                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`15551556                    Permission policy for tool execution.15571558                    - `BetaManagedAgentsAlwaysAllowPolicy object`15591560                      Tool calls are automatically approved without user confirmation.15611562                    - `BetaManagedAgentsAlwaysAskPolicy object`15631564                      Tool calls require user confirmation before execution.15651566                    - `BetaManagedAgentsAutoPolicy object`15671568                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.15691570                  - `allowed_domains: optional array of string`15711572                  - `blocked_domains: optional array of string`15731574                  - `user_location: optional BetaManagedAgentsUserLocation or null`15751576                    Approximate user location for search result localization.15771578                    - `type: "approximate"`15791580                      Location precision. Only "approximate" is supported.15811582                    - `city: optional string or null`15831584                      City name.15851586                      minLength: 1, maxLength: 25515871588                    - `country: optional string or null`15891590                      Two-letter ISO 3166-1 country code, uppercase.15911592                    - `region: optional string or null`15931594                      Region or state name.15951596                      minLength: 1, maxLength: 25515971598                    - `timezone: optional string or null`15991600                      IANA timezone identifier, e.g. "America/Los_Angeles".16011602                      minLength: 1, maxLength: 25516031604              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`16051606                Resolved default configuration for agent tools.16071608                - `enabled: boolean`16091610                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`16111612                  Permission policy for tool execution.16131614                  - `BetaManagedAgentsAlwaysAllowPolicy object`16151616                    Tool calls are automatically approved without user confirmation.16171618                  - `BetaManagedAgentsAlwaysAskPolicy object`16191620                    Tool calls require user confirmation before execution.16211622                  - `BetaManagedAgentsAutoPolicy object`16231624                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.16251626            - `BetaManagedAgentsMCPToolset object`16271628              - `type: "mcp_toolset"`16291630              - `configs: array of BetaManagedAgentsMCPToolConfig`16311632                - `enabled: boolean`16331634                - `name: string`16351636                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`16371638                  Permission policy for tool execution.16391640                  - `BetaManagedAgentsAlwaysAllowPolicy object`16411642                    Tool calls are automatically approved without user confirmation.16431644                  - `BetaManagedAgentsAlwaysAskPolicy object`16451646                    Tool calls require user confirmation before execution.16471648                  - `BetaManagedAgentsAutoPolicy object`16491650                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.16511652              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`16531654                Resolved default configuration for all tools from an MCP server.16551656                - `enabled: boolean`16571658                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`16591660                  Permission policy for tool execution.16611662                  - `BetaManagedAgentsAlwaysAllowPolicy object`16631664                    Tool calls are automatically approved without user confirmation.16651666                  - `BetaManagedAgentsAlwaysAskPolicy object`16671668                    Tool calls require user confirmation before execution.16691670                  - `BetaManagedAgentsAutoPolicy object`16711672                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.16731674              - `mcp_server_name: string`16751676            - `BetaManagedAgentsCustomTool object`16771678              A custom tool as returned in API responses.16791680              - `type: "custom"`16811682              - `description: string`16831684              - `input_schema: BetaManagedAgentsCustomToolInputSchema`16851686                JSON Schema for custom tool input parameters.16871688                - `type: "object"`16891690                - `properties: optional map[unknown] or null`16911692                - `required: optional array of string or null`16931694              - `name: string`16951696          - `version: number`16971698            format: int3216991700        - `BetaManagedAgentsAdvisor object`17011702          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.17031704          - `type: "advisor"`17051706          - `model: string`17071708            The advisor model id.17091710    - `name: string`17111712    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`17131714      - `BetaManagedAgentsAnthropicSkill object`17151716        A resolved Anthropic-managed skill.17171718      - `BetaManagedAgentsCustomSkill object`17191720        A resolved user-created custom skill.17211722    - `system: string or null`17231724    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`17251726      - `BetaManagedAgentsAgentToolset20260401 object`17271728      - `BetaManagedAgentsMCPToolset object`17291730      - `BetaManagedAgentsCustomTool object`17311732        A custom tool as returned in API responses.17331734    - `version: number`17351736      format: int3217371738  - `archived_at: string or null`17391740    A timestamp in RFC 3339 format17411742    format: date-time17431744  - `budget: BetaManagedAgentsBudgetLimit or null`17451746    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.17471748    - `type: "limit"`17491750    - `max_list_cost: BetaMonetaryAmount`17511752      A monetary amount in a specific currency.17531754      - `amount: string`17551756        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.17571758      - `currency: BetaCurrency`17591760        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.17611762  - `created_at: string`17631764    A timestamp in RFC 3339 format17651766    format: date-time17671768  - `environment_id: string`17691770  - `metadata: map[string]`17711772  - `outcome_evaluations: array of BetaManagedAgentsOutcomeEvaluationResource`17731774    Per-outcome evaluation state. One entry per `define_outcome` event sent to the session.17751776    - `type: "outcome_evaluation"`17771778    - `completed_at: string or null`17791780      A timestamp in RFC 3339 format17811782      format: date-time17831784    - `description: string`17851786      What the agent should produce.17871788    - `explanation: string or null`17891790      Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.17911792    - `iteration: number`17931794      0-indexed revision cycle the outcome is currently on.17951796      format: int3217971798    - `outcome_id: string`17991800      Server-generated outc_ ID for this outcome.18011802    - `result: string`18031804      Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.18051806  - `resources: array of BetaManagedAgentsSessionResource`18071808    - `BetaManagedAgentsGitHubRepositoryResource object`18091810      - `type: "github_repository"`18111812      - `id: string`18131814      - `created_at: string`18151816        A timestamp in RFC 3339 format18171818        format: date-time18191820      - `mount_path: string`18211822      - `updated_at: string`18231824        A timestamp in RFC 3339 format18251826        format: date-time18271828      - `url: string`18291830      - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`18311832        - `BetaManagedAgentsBranchCheckout object`18331834          - `type: "branch"`18351836          - `name: string`18371838            Branch name to check out.18391840            minLength: 1, maxLength: 25518411842        - `BetaManagedAgentsCommitCheckout object`18431844          - `type: "commit"`18451846          - `sha: string`18471848            Full commit SHA to check out.18491850            minLength: 7, maxLength: 6418511852    - `BetaManagedAgentsFileResource object`18531854      - `type: "file"`18551856      - `id: string`18571858      - `created_at: string`18591860        A timestamp in RFC 3339 format18611862        format: date-time18631864      - `file_id: string`18651866      - `mount_path: string`18671868      - `updated_at: string`18691870        A timestamp in RFC 3339 format18711872        format: date-time18731874    - `BetaManagedAgentsMemoryStoreResource object`18751876      A memory store attached to an agent session.18771878      - `type: "memory_store"`18791880      - `memory_store_id: string`18811882        The memory store ID (memstore_...). Must belong to the caller's organization and workspace.18831884      - `access: optional "read_write" or "read_only" or null`18851886        Access mode for an attached memory store.18871888        - `"read_write"`18891890        - `"read_only"`18911892      - `description: optional string`18931894        Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.18951896      - `instructions: optional string or null`18971898        Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.18991900        maxLength: 409619011902      - `mount_path: optional string or null`19031904        Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.19051906      - `name: optional string or null`19071908        Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.19091910  - `stats: BetaManagedAgentsSessionStats`19111912    Timing statistics for a session.19131914    - `active_seconds: optional number`19151916      Cumulative time in seconds the session spent in `running` status. Excludes idle time.19171918      format: double19191920    - `duration_seconds: optional number`19211922      Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.19231924      format: double19251926  - `status: "rescheduling" or "running" or "idle" or "terminated"`19271928    SessionStatus enum19291930    - `"rescheduling"`19311932    - `"running"`19331934    - `"idle"`19351936    - `"terminated"`19371938  - `title: string or null`19391940  - `updated_at: string`19411942    A timestamp in RFC 3339 format19431944    format: date-time19451946  - `usage: BetaManagedAgentsSessionUsage`19471948    Cumulative token usage for a session across all turns.19491950    - `active_seconds: optional number`19511952      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.19531954      format: double19551956    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`19571958      Prompt-cache creation token usage broken down by cache lifetime.19591960      - `ephemeral_1h_input_tokens: optional number`19611962        Tokens used to create 1-hour ephemeral cache entries.19631964        format: int3219651966      - `ephemeral_5m_input_tokens: optional number`19671968        Tokens used to create 5-minute ephemeral cache entries.19691970        format: int3219711972    - `cache_read_input_tokens: optional number`19731974      Total tokens read from prompt cache.19751976      format: int3219771978    - `input_tokens: optional number`19791980      Total input tokens consumed across all turns.19811982      format: int3219831984    - `list_cost: optional BetaMonetaryAmount or null`19851986      A monetary amount in a specific currency.19871988    - `output_tokens: optional number`19891990      Total output tokens generated across all turns.19911992      format: int3219931994    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`19951996      Cumulative count of server-executed tool invocations, broken down by tool.19971998      - `web_fetch_requests: optional number`19992000        Number of server-executed web fetch requests.20012002        format: int3220032004      - `web_search_requests: optional number`20052006        Number of server-executed web search requests.20072008        format: int3220092010  - `vault_ids: array of string`20112012    Vault IDs attached to the session at creation. Empty when no vaults were supplied.20132014  - `deployment_id: optional string or null`20152016    Deployment ID when the session was created from a deployment reference. Null otherwise.20172018### Example20192020```bash2021curl https://api.anthropic.com/v1/sessions \2022    -H 'Content-Type: application/json' \2023    -H 'anthropic-version: 2023-06-01' \2024    -H 'anthropic-beta: managed-agents-2026-04-01' \2025    -H "X-Api-Key: $ANTHROPIC_API_KEY" \2026    -d '{2027          "agent": "agent_011CZkYpogX7uDKUyvBTophP",2028          "environment_id": "env_011CZkZ9X2dpNyB7HsEFoRfW",2029          "title": "Order #1234 inquiry"2030        }'2031```20322033#### Response (200)20342035```json2036{2037  "id": "sesn_011CZkZAtmR3yMPDzynEDxu7",2038  "agent": {2039    "id": "agent_011CZkYpogX7uDKUyvBTophP",2040    "description": "A general-purpose starter agent.",2041    "mcp_servers": [2042      {2043        "name": "example-mcp",2044        "type": "url",2045        "url": "https://example-server.modelcontextprotocol.io/sse"2046      }2047    ],2048    "model": {2049      "id": "claude-opus-5",2050      "effort": {2051        "type": "low"2052      },2053      "inference_geo": "inference_geo",2054      "speed": "standard"2055    },2056    "multiagent": {2057      "agents": [2058        {2059          "id": "agent_011CZkYqphY8vELVzwCUpqiQ",2060          "description": "A focused research subagent.",2061          "mcp_servers": [2062            {2063              "name": "example-mcp",2064              "type": "url",2065              "url": "https://example-server.modelcontextprotocol.io/sse"2066            }2067          ],2068          "model": {2069            "id": "claude-opus-5",2070            "effort": {2071              "type": "low"2072            },2073            "inference_geo": "inference_geo",2074            "speed": "standard"2075          },2076          "name": "Researcher",2077          "skills": [2078            {2079              "skill_id": "xlsx",2080              "type": "anthropic",2081              "version": "1"2082            }2083          ],2084          "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",2085          "tools": [2086            {2087              "configs": [2088                {2089                  "enabled": true,2090                  "name": "bash",2091                  "permission_policy": {2092                    "type": "always_allow"2093                  },2094                  "type": "bash"2095                }2096              ],2097              "default_config": {2098                "enabled": true,2099                "permission_policy": {2100                  "type": "always_ask"2101                }2102              },2103              "type": "agent_toolset_20260401"2104            }2105          ],2106          "type": "agent",2107          "version": 12108        }2109      ],2110      "type": "coordinator"2111    },2112    "name": "My First Agent",2113    "skills": [2114      {2115        "skill_id": "xlsx",2116        "type": "anthropic",2117        "version": "1"2118      },2119      {2120        "skill_id": "skill_011CZkZFNu9hAbo3jZPRgTlx",2121        "type": "custom",2122        "version": "2"2123      }2124    ],2125    "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user's task end to end.",2126    "tools": [2127      {2128        "configs": [2129          {2130            "enabled": true,2131            "name": "bash",2132            "permission_policy": {2133              "type": "always_allow"2134            },2135            "type": "bash"2136          }2137        ],2138        "default_config": {2139          "enabled": true,2140          "permission_policy": {2141            "type": "always_ask"2142          }2143        },2144        "type": "agent_toolset_20260401"2145      }2146    ],2147    "type": "agent",2148    "version": 12149  },2150  "archived_at": null,2151  "budget": {2152    "max_list_cost": {2153      "amount": "2500",2154      "currency": "USD"2155    },2156    "type": "limit"2157  },2158  "created_at": "2026-03-15T10:00:00Z",2159  "environment_id": "env_011CZkZ9X2dpNyB7HsEFoRfW",2160  "metadata": {},2161  "outcome_evaluations": [2162    {2163      "completed_at": "2026-03-15T10:02:31Z",2164      "description": "Produce a 2-page summary as summary.md",2165      "explanation": "All five sections present with inline citations.",2166      "iteration": 0,2167      "outcome_id": "outc_011CZkZRSw2kEfs6ncTVljxP",2168      "result": "satisfied",2169      "type": "outcome_evaluation"2170    }2171  ],2172  "resources": [2173    {2174      "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",2175      "created_at": "2026-03-15T10:00:00Z",2176      "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",2177      "mount_path": "/uploads/receipt.pdf",2178      "type": "file",2179      "updated_at": "2026-03-15T10:00:00Z"2180    },2181    {2182      "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",2183      "created_at": "2026-03-15T10:00:00Z",2184      "mount_path": "/workspace/example-repo",2185      "type": "github_repository",2186      "updated_at": "2026-03-15T10:00:00Z",2187      "url": "https://github.com/example-org/example-repo",2188      "checkout": {2189        "name": "main",2190        "type": "branch"2191      }2192    }2193  ],2194  "stats": {2195    "active_seconds": 0,2196    "duration_seconds": 02197  },2198  "status": "idle",2199  "title": "Order #1234 inquiry",2200  "type": "session",2201  "updated_at": "2026-03-15T10:00:00Z",2202  "usage": {2203    "active_seconds": 0,2204    "cache_creation": {2205      "ephemeral_1h_input_tokens": 0,2206      "ephemeral_5m_input_tokens": 02207    },2208    "cache_read_input_tokens": 0,2209    "input_tokens": 0,2210    "list_cost": {2211      "amount": "2500",2212      "currency": "USD"2213    },2214    "output_tokens": 0,2215    "server_tool_use": {2216      "web_fetch_requests": 0,2217      "web_search_requests": 32218    }2219  },2220  "vault_ids": [2221    "vlt_011CZkZDLs7fYzm1hXNPeRjv"2222  ],2223  "deployment_id": "deployment_id"2224}2225```22262227## List Sessions22282229**GET** `/v1/sessions`22302231List Sessions22322233### Query parameters22342235- `agent_id: optional string`22362237  Filter sessions created with this agent ID.22382239- `agent_version: optional number`22402241  Filter by agent version. Only applies when `agent_id` is also set.22422243  format: int3222442245- `"created_at[gt]": optional string`22462247  Return sessions created after this time (exclusive).22482249  format: date-time22502251- `"created_at[gte]": optional string`22522253  Return sessions created at or after this time (inclusive).22542255  format: date-time22562257- `"created_at[lt]": optional string`22582259  Return sessions created before this time (exclusive).22602261  format: date-time22622263- `"created_at[lte]": optional string`22642265  Return sessions created at or before this time (inclusive).22662267  format: date-time22682269- `deployment_id: optional string`22702271  Filter sessions created by this deployment ID.22722273- `include_archived: optional boolean`22742275  When true, includes archived sessions. Default: false (exclude archived).22762277- `limit: optional number`22782279  Maximum number of results to return.22802281  format: int3222822283- `memory_store_id: optional string`22842285  Filter sessions whose resources contain a `memory_store` with this memory store ID.22862287- `order: optional "asc" or "desc"`22882289  Sort direction for results, ordered by `created_at`. Defaults to `desc` (newest first).22902291  - `"asc"`22922293  - `"desc"`22942295- `page: optional string`22962297  Opaque pagination cursor from a previous response.22982299- `statuses: optional array of "rescheduling" or "running" or "idle" or "terminated"`23002301  Filter by session status. Repeat the parameter to match any of multiple statuses.23022303  - `"rescheduling"`23042305  - `"running"`23062307  - `"idle"`23082309  - `"terminated"`23102311### Headers23122313- `"anthropic-beta": optional array of AnthropicBeta`23142315  Optional header to specify the beta version(s) you want to use.23162317  - `string`23182319  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`23202321    - `"message-batches-2024-09-24"`23222323    - `"prompt-caching-2024-07-31"`23242325    - `"computer-use-2024-10-22"`23262327    - `"computer-use-2025-01-24"`23282329    - `"pdfs-2024-09-25"`23302331    - `"token-counting-2024-11-01"`23322333    - `"token-efficient-tools-2025-02-19"`23342335    - `"output-128k-2025-02-19"`23362337    - `"files-api-2025-04-14"`23382339    - `"mcp-client-2025-04-04"`23402341    - `"mcp-client-2025-11-20"`23422343    - `"dev-full-thinking-2025-05-14"`23442345    - `"interleaved-thinking-2025-05-14"`23462347    - `"code-execution-2025-05-22"`23482349    - `"extended-cache-ttl-2025-04-11"`23502351    - `"context-1m-2025-08-07"`23522353    - `"context-management-2025-06-27"`23542355    - `"model-context-window-exceeded-2025-08-26"`23562357    - `"skills-2025-10-02"`23582359    - `"fast-mode-2026-02-01"`23602361    - `"output-300k-2026-03-24"`23622363    - `"user-profiles-2026-03-24"`23642365    - `"user-profiles-2026-08-18"`23662367    - `"user-profiles-2026-09-04"`23682369    - `"advisor-tool-2026-03-01"`23702371    - `"managed-agents-2026-04-01"`23722373    - `"cache-diagnosis-2026-04-07"`23742375    - `"dreaming-2026-04-21"`23762377    - `"thinking-token-count-2026-05-13"`23782379    - `"server-side-fallback-2026-06-01"`23802381    - `"server-side-fallback-2026-07-01"`23822383    - `"fallback-credit-2026-06-01"`23842385    - `"fallback-credit-2026-07-01"`23862387    - `"agent-memory-2026-07-22"`23882389    - `"mid-conversation-tool-changes-2026-07-01"`23902391    - `"compact-2026-01-12"`23922393    - `"computer-use-2025-11-24"`23942395    - `"mcp-tunnels-2026-06-22"`23962397    - `"structured-outputs-2025-11-13"`23982399    - `"task-budgets-2026-03-13"`24002401    - `"thinking-display-updates-2026-08-18"`24022403    - `"ce-user-management-2026-07-13"`24042405    - `"mid-conversation-output-config-2026-07-01"`24062407    - `"thinking-binding-controls-2026-08-01"`24082409    - `"mid-conversation-system-clear-at-2026-08-21"`24102411    - `"compact-2026-09-04"`24122413- `"anthropic-workspace-id": optional string`24142415### Returns24162417- `data: optional array of BetaManagedAgentsSession`24182419  List of sessions.24202421  - `type: "session"`24222423  - `id: string`24242425  - `agent: BetaManagedAgentsSessionAgent`24262427    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.24282429    - `type: "agent"`24302431    - `id: string`24322433    - `description: string or null`24342435    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`24362437      - `type: "url"`24382439      - `name: string`24402441      - `url: string`24422443    - `model: BetaManagedAgentsModelConfig`24442445      Model identifier and configuration.24462447      - `id: BetaManagedAgentsModel`24482449        The model that will power your agent.24502451        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.24522453        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`24542455          The model that will power your agent.24562457          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.24582459          - `"claude-fable-5-1"`24602461            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows24622463          - `"claude-sonnet-5"`24642465            High-performance model for coding and agents24662467          - `"claude-fable-5"`24682469            Next generation of intelligence for the hardest knowledge work and coding problems24702471          - `"claude-opus-5"`24722473            Powerful intelligence for long-running agents and coding24742475          - `"claude-opus-4-8"`24762477            Powerful intelligence for long-running agents and coding24782479          - `"claude-opus-4-7"`24802481            Powerful intelligence for long-running agents and coding24822483          - `"claude-opus-4-6"`24842485            Powerful intelligence for long-running agents and coding24862487          - `"claude-sonnet-4-6"`24882489            Best combination of speed and intelligence24902491          - `"claude-haiku-4-5"`24922493            Fastest model with near-frontier intelligence24942495          - `"claude-haiku-4-5-20251001"`24962497            Fastest model with near-frontier intelligence24982499          - `"claude-opus-4-5"`25002501            Powerful intelligence for long-running agents and coding25022503          - `"claude-opus-4-5-20251101"`25042505            Powerful intelligence for long-running agents and coding25062507          - `"claude-sonnet-4-5"`25082509            High-performance model for agents and coding25102511          - `"claude-sonnet-4-5-20250929"`25122513            High-performance model for agents and coding25142515        - `string`25162517      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`25182519        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.25202521        - `BetaManagedAgentsEffortLow object`25222523          Low effort. Favors latency over reasoning depth.25242525          - `type: "low"`25262527        - `BetaManagedAgentsEffortMedium object`25282529          Medium effort. Balances latency and reasoning depth.25302531          - `type: "medium"`25322533        - `BetaManagedAgentsEffortHigh object`25342535          High effort. Favors reasoning depth.25362537          - `type: "high"`25382539        - `BetaManagedAgentsEffortXhigh object`25402541          Extra-high effort. Not all models accept this level.25422543          - `type: "xhigh"`25442545        - `BetaManagedAgentsEffortMax object`25462547          Maximum effort. Favors reasoning depth over latency.25482549          - `type: "max"`25502551      - `inference_geo: optional string`25522553        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.25542555      - `speed: optional "standard" or "fast"`25562557        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.25582559        - `"standard"`25602561        - `"fast"`25622563    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`25642565      Resolved coordinator topology with full agent definitions for each roster member.25662567      - `type: "coordinator"`25682569      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`25702571        Full `agent` definitions the coordinator may spawn as session threads.25722573        - `BetaManagedAgentsSessionThreadAgent object`25742575          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.25762577          - `type: "agent"`25782579          - `id: string`25802581          - `description: string or null`25822583          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`25842585            - `type: "url"`25862587            - `name: string`25882589            - `url: string`25902591          - `model: BetaManagedAgentsModelConfig`25922593            Model identifier and configuration.25942595          - `name: string`25962597          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`25982599            - `BetaManagedAgentsAnthropicSkill object`26002601              A resolved Anthropic-managed skill.26022603              - `type: "anthropic"`26042605              - `skill_id: string`26062607              - `version: string`26082609            - `BetaManagedAgentsCustomSkill object`26102611              A resolved user-created custom skill.26122613              - `type: "custom"`26142615              - `skill_id: string`26162617              - `version: string`26182619          - `system: string or null`26202621          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`26222623            - `BetaManagedAgentsAgentToolset20260401 object`26242625              - `type: "agent_toolset_20260401"`26262627              - `configs: array of BetaManagedAgentsAgentToolConfig`26282629                - `BetaManagedAgentsBashToolConfig object`26302631                  Configuration for the bash tool.26322633                  - `type: "bash"`26342635                  - `enabled: boolean`26362637                  - `name: "bash"`26382639                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`26402641                    Permission policy for tool execution.26422643                    - `BetaManagedAgentsAlwaysAllowPolicy object`26442645                      Tool calls are automatically approved without user confirmation.26462647                      - `type: "always_allow"`26482649                    - `BetaManagedAgentsAlwaysAskPolicy object`26502651                      Tool calls require user confirmation before execution.26522653                      - `type: "always_ask"`26542655                    - `BetaManagedAgentsAutoPolicy object`26562657                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.26582659                      - `type: "auto"`26602661                - `BetaManagedAgentsEditToolConfig object`26622663                  Configuration for the edit tool.26642665                  - `type: "edit"`26662667                  - `enabled: boolean`26682669                  - `name: "edit"`26702671                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`26722673                    Permission policy for tool execution.26742675                    - `BetaManagedAgentsAlwaysAllowPolicy object`26762677                      Tool calls are automatically approved without user confirmation.26782679                    - `BetaManagedAgentsAlwaysAskPolicy object`26802681                      Tool calls require user confirmation before execution.26822683                    - `BetaManagedAgentsAutoPolicy object`26842685                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.26862687                - `BetaManagedAgentsReadToolConfig object`26882689                  Configuration for the read tool.26902691                  - `type: "read"`26922693                  - `enabled: boolean`26942695                  - `name: "read"`26962697                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`26982699                    Permission policy for tool execution.27002701                    - `BetaManagedAgentsAlwaysAllowPolicy object`27022703                      Tool calls are automatically approved without user confirmation.27042705                    - `BetaManagedAgentsAlwaysAskPolicy object`27062707                      Tool calls require user confirmation before execution.27082709                    - `BetaManagedAgentsAutoPolicy object`27102711                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.27122713                - `BetaManagedAgentsWriteToolConfig object`27142715                  Configuration for the write tool.27162717                  - `type: "write"`27182719                  - `enabled: boolean`27202721                  - `name: "write"`27222723                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`27242725                    Permission policy for tool execution.27262727                    - `BetaManagedAgentsAlwaysAllowPolicy object`27282729                      Tool calls are automatically approved without user confirmation.27302731                    - `BetaManagedAgentsAlwaysAskPolicy object`27322733                      Tool calls require user confirmation before execution.27342735                    - `BetaManagedAgentsAutoPolicy object`27362737                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.27382739                - `BetaManagedAgentsGlobToolConfig object`27402741                  Configuration for the glob tool.27422743                  - `type: "glob"`27442745                  - `enabled: boolean`27462747                  - `name: "glob"`27482749                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`27502751                    Permission policy for tool execution.27522753                    - `BetaManagedAgentsAlwaysAllowPolicy object`27542755                      Tool calls are automatically approved without user confirmation.27562757                    - `BetaManagedAgentsAlwaysAskPolicy object`27582759                      Tool calls require user confirmation before execution.27602761                    - `BetaManagedAgentsAutoPolicy object`27622763                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.27642765                - `BetaManagedAgentsGrepToolConfig object`27662767                  Configuration for the grep tool.27682769                  - `type: "grep"`27702771                  - `enabled: boolean`27722773                  - `name: "grep"`27742775                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`27762777                    Permission policy for tool execution.27782779                    - `BetaManagedAgentsAlwaysAllowPolicy object`27802781                      Tool calls are automatically approved without user confirmation.27822783                    - `BetaManagedAgentsAlwaysAskPolicy object`27842785                      Tool calls require user confirmation before execution.27862787                    - `BetaManagedAgentsAutoPolicy object`27882789                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.27902791                - `BetaManagedAgentsWebFetchToolConfig object`27922793                  Configuration for the web_fetch tool.27942795                  - `type: "web_fetch"`27962797                  - `enabled: boolean`27982799                  - `name: "web_fetch"`28002801                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`28022803                    Permission policy for tool execution.28042805                    - `BetaManagedAgentsAlwaysAllowPolicy object`28062807                      Tool calls are automatically approved without user confirmation.28082809                    - `BetaManagedAgentsAlwaysAskPolicy object`28102811                      Tool calls require user confirmation before execution.28122813                    - `BetaManagedAgentsAutoPolicy object`28142815                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.28162817                  - `allowed_domains: optional array of string`28182819                  - `blocked_domains: optional array of string`28202821                  - `max_content_tokens: optional number or null`28222823                    format: int3228242825                - `BetaManagedAgentsWebSearchToolConfig object`28262827                  Configuration for the web_search tool.28282829                  - `type: "web_search"`28302831                  - `enabled: boolean`28322833                  - `name: "web_search"`28342835                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`28362837                    Permission policy for tool execution.28382839                    - `BetaManagedAgentsAlwaysAllowPolicy object`28402841                      Tool calls are automatically approved without user confirmation.28422843                    - `BetaManagedAgentsAlwaysAskPolicy object`28442845                      Tool calls require user confirmation before execution.28462847                    - `BetaManagedAgentsAutoPolicy object`28482849                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.28502851                  - `allowed_domains: optional array of string`28522853                  - `blocked_domains: optional array of string`28542855                  - `user_location: optional BetaManagedAgentsUserLocation or null`28562857                    Approximate user location for search result localization.28582859                    - `type: "approximate"`28602861                      Location precision. Only "approximate" is supported.28622863                    - `city: optional string or null`28642865                      City name.28662867                      minLength: 1, maxLength: 25528682869                    - `country: optional string or null`28702871                      Two-letter ISO 3166-1 country code, uppercase.28722873                    - `region: optional string or null`28742875                      Region or state name.28762877                      minLength: 1, maxLength: 25528782879                    - `timezone: optional string or null`28802881                      IANA timezone identifier, e.g. "America/Los_Angeles".28822883                      minLength: 1, maxLength: 25528842885              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`28862887                Resolved default configuration for agent tools.28882889                - `enabled: boolean`28902891                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`28922893                  Permission policy for tool execution.28942895                  - `BetaManagedAgentsAlwaysAllowPolicy object`28962897                    Tool calls are automatically approved without user confirmation.28982899                  - `BetaManagedAgentsAlwaysAskPolicy object`29002901                    Tool calls require user confirmation before execution.29022903                  - `BetaManagedAgentsAutoPolicy object`29042905                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.29062907            - `BetaManagedAgentsMCPToolset object`29082909              - `type: "mcp_toolset"`29102911              - `configs: array of BetaManagedAgentsMCPToolConfig`29122913                - `enabled: boolean`29142915                - `name: string`29162917                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`29182919                  Permission policy for tool execution.29202921                  - `BetaManagedAgentsAlwaysAllowPolicy object`29222923                    Tool calls are automatically approved without user confirmation.29242925                  - `BetaManagedAgentsAlwaysAskPolicy object`29262927                    Tool calls require user confirmation before execution.29282929                  - `BetaManagedAgentsAutoPolicy object`29302931                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.29322933              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`29342935                Resolved default configuration for all tools from an MCP server.29362937                - `enabled: boolean`29382939                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`29402941                  Permission policy for tool execution.29422943                  - `BetaManagedAgentsAlwaysAllowPolicy object`29442945                    Tool calls are automatically approved without user confirmation.29462947                  - `BetaManagedAgentsAlwaysAskPolicy object`29482949                    Tool calls require user confirmation before execution.29502951                  - `BetaManagedAgentsAutoPolicy object`29522953                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.29542955              - `mcp_server_name: string`29562957            - `BetaManagedAgentsCustomTool object`29582959              A custom tool as returned in API responses.29602961              - `type: "custom"`29622963              - `description: string`29642965              - `input_schema: BetaManagedAgentsCustomToolInputSchema`29662967                JSON Schema for custom tool input parameters.29682969                - `type: "object"`29702971                - `properties: optional map[unknown] or null`29722973                - `required: optional array of string or null`29742975              - `name: string`29762977          - `version: number`29782979            format: int3229802981        - `BetaManagedAgentsAdvisor object`29822983          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.29842985          - `type: "advisor"`29862987          - `model: string`29882989            The advisor model id.29902991    - `name: string`29922993    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`29942995      - `BetaManagedAgentsAnthropicSkill object`29962997        A resolved Anthropic-managed skill.29982999      - `BetaManagedAgentsCustomSkill object`30003001        A resolved user-created custom skill.30023003    - `system: string or null`30043005    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`30063007      - `BetaManagedAgentsAgentToolset20260401 object`30083009      - `BetaManagedAgentsMCPToolset object`30103011      - `BetaManagedAgentsCustomTool object`30123013        A custom tool as returned in API responses.30143015    - `version: number`30163017      format: int3230183019  - `archived_at: string or null`30203021    A timestamp in RFC 3339 format30223023    format: date-time30243025  - `budget: BetaManagedAgentsBudgetLimit or null`30263027    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.30283029    - `type: "limit"`30303031    - `max_list_cost: BetaMonetaryAmount`30323033      A monetary amount in a specific currency.30343035      - `amount: string`30363037        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.30383039      - `currency: BetaCurrency`30403041        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.30423043  - `created_at: string`30443045    A timestamp in RFC 3339 format30463047    format: date-time30483049  - `environment_id: string`30503051  - `metadata: map[string]`30523053  - `outcome_evaluations: array of BetaManagedAgentsOutcomeEvaluationResource`30543055    Per-outcome evaluation state. One entry per `define_outcome` event sent to the session.30563057    - `type: "outcome_evaluation"`30583059    - `completed_at: string or null`30603061      A timestamp in RFC 3339 format30623063      format: date-time30643065    - `description: string`30663067      What the agent should produce.30683069    - `explanation: string or null`30703071      Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.30723073    - `iteration: number`30743075      0-indexed revision cycle the outcome is currently on.30763077      format: int3230783079    - `outcome_id: string`30803081      Server-generated outc_ ID for this outcome.30823083    - `result: string`30843085      Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.30863087  - `resources: array of BetaManagedAgentsSessionResource`30883089    - `BetaManagedAgentsGitHubRepositoryResource object`30903091      - `type: "github_repository"`30923093      - `id: string`30943095      - `created_at: string`30963097        A timestamp in RFC 3339 format30983099        format: date-time31003101      - `mount_path: string`31023103      - `updated_at: string`31043105        A timestamp in RFC 3339 format31063107        format: date-time31083109      - `url: string`31103111      - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`31123113        - `BetaManagedAgentsBranchCheckout object`31143115          - `type: "branch"`31163117          - `name: string`31183119            Branch name to check out.31203121            minLength: 1, maxLength: 25531223123        - `BetaManagedAgentsCommitCheckout object`31243125          - `type: "commit"`31263127          - `sha: string`31283129            Full commit SHA to check out.31303131            minLength: 7, maxLength: 6431323133    - `BetaManagedAgentsFileResource object`31343135      - `type: "file"`31363137      - `id: string`31383139      - `created_at: string`31403141        A timestamp in RFC 3339 format31423143        format: date-time31443145      - `file_id: string`31463147      - `mount_path: string`31483149      - `updated_at: string`31503151        A timestamp in RFC 3339 format31523153        format: date-time31543155    - `BetaManagedAgentsMemoryStoreResource object`31563157      A memory store attached to an agent session.31583159      - `type: "memory_store"`31603161      - `memory_store_id: string`31623163        The memory store ID (memstore_...). Must belong to the caller's organization and workspace.31643165      - `access: optional "read_write" or "read_only" or null`31663167        Access mode for an attached memory store.31683169        - `"read_write"`31703171        - `"read_only"`31723173      - `description: optional string`31743175        Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.31763177      - `instructions: optional string or null`31783179        Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.31803181        maxLength: 409631823183      - `mount_path: optional string or null`31843185        Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.31863187      - `name: optional string or null`31883189        Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.31903191  - `stats: BetaManagedAgentsSessionStats`31923193    Timing statistics for a session.31943195    - `active_seconds: optional number`31963197      Cumulative time in seconds the session spent in `running` status. Excludes idle time.31983199      format: double32003201    - `duration_seconds: optional number`32023203      Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.32043205      format: double32063207  - `status: "rescheduling" or "running" or "idle" or "terminated"`32083209    SessionStatus enum32103211    - `"rescheduling"`32123213    - `"running"`32143215    - `"idle"`32163217    - `"terminated"`32183219  - `title: string or null`32203221  - `updated_at: string`32223223    A timestamp in RFC 3339 format32243225    format: date-time32263227  - `usage: BetaManagedAgentsSessionUsage`32283229    Cumulative token usage for a session across all turns.32303231    - `active_seconds: optional number`32323233      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.32343235      format: double32363237    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`32383239      Prompt-cache creation token usage broken down by cache lifetime.32403241      - `ephemeral_1h_input_tokens: optional number`32423243        Tokens used to create 1-hour ephemeral cache entries.32443245        format: int3232463247      - `ephemeral_5m_input_tokens: optional number`32483249        Tokens used to create 5-minute ephemeral cache entries.32503251        format: int3232523253    - `cache_read_input_tokens: optional number`32543255      Total tokens read from prompt cache.32563257      format: int3232583259    - `input_tokens: optional number`32603261      Total input tokens consumed across all turns.32623263      format: int3232643265    - `list_cost: optional BetaMonetaryAmount or null`32663267      A monetary amount in a specific currency.32683269    - `output_tokens: optional number`32703271      Total output tokens generated across all turns.32723273      format: int3232743275    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`32763277      Cumulative count of server-executed tool invocations, broken down by tool.32783279      - `web_fetch_requests: optional number`32803281        Number of server-executed web fetch requests.32823283        format: int3232843285      - `web_search_requests: optional number`32863287        Number of server-executed web search requests.32883289        format: int3232903291  - `vault_ids: array of string`32923293    Vault IDs attached to the session at creation. Empty when no vaults were supplied.32943295  - `deployment_id: optional string or null`32963297    Deployment ID when the session was created from a deployment reference. Null otherwise.32983299- `next_page: optional string or null`33003301  Opaque cursor for the next page. Null when no more results.33023303- `prev_page: optional string or null`33043305  Opaque cursor for the previous page. Null when on the first page. Pass as the `page` parameter to navigate backward.33063307### Example33083309```bash3310curl https://api.anthropic.com/v1/sessions \3311    -H 'anthropic-version: 2023-06-01' \3312    -H 'anthropic-beta: managed-agents-2026-04-01' \3313    -H "X-Api-Key: $ANTHROPIC_API_KEY"3314```33153316#### Response (200)33173318```json3319{3320  "data": [3321    {3322      "id": "sesn_011CZkZAtmR3yMPDzynEDxu7",3323      "agent": {3324        "id": "agent_011CZkYpogX7uDKUyvBTophP",3325        "description": "A general-purpose starter agent.",3326        "mcp_servers": [3327          {3328            "name": "example-mcp",3329            "type": "url",3330            "url": "https://example-server.modelcontextprotocol.io/sse"3331          }3332        ],3333        "model": {3334          "id": "claude-opus-5",3335          "effort": {3336            "type": "low"3337          },3338          "inference_geo": "inference_geo",3339          "speed": "standard"3340        },3341        "multiagent": {3342          "agents": [3343            {3344              "id": "agent_011CZkYqphY8vELVzwCUpqiQ",3345              "description": "A focused research subagent.",3346              "mcp_servers": [3347                {3348                  "name": "example-mcp",3349                  "type": "url",3350                  "url": "https://example-server.modelcontextprotocol.io/sse"3351                }3352              ],3353              "model": {3354                "id": "claude-opus-5",3355                "effort": {3356                  "type": "low"3357                },3358                "inference_geo": "inference_geo",3359                "speed": "standard"3360              },3361              "name": "Researcher",3362              "skills": [3363                {3364                  "skill_id": "xlsx",3365                  "type": "anthropic",3366                  "version": "1"3367                }3368              ],3369              "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",3370              "tools": [3371                {3372                  "configs": [3373                    {3374                      "enabled": true,3375                      "name": "bash",3376                      "permission_policy": {3377                        "type": "always_allow"3378                      },3379                      "type": "bash"3380                    }3381                  ],3382                  "default_config": {3383                    "enabled": true,3384                    "permission_policy": {3385                      "type": "always_ask"3386                    }3387                  },3388                  "type": "agent_toolset_20260401"3389                }3390              ],3391              "type": "agent",3392              "version": 13393            }3394          ],3395          "type": "coordinator"3396        },3397        "name": "My First Agent",3398        "skills": [3399          {3400            "skill_id": "xlsx",3401            "type": "anthropic",3402            "version": "1"3403          },3404          {3405            "skill_id": "skill_011CZkZFNu9hAbo3jZPRgTlx",3406            "type": "custom",3407            "version": "2"3408          }3409        ],3410        "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user's task end to end.",3411        "tools": [3412          {3413            "configs": [3414              {3415                "enabled": true,3416                "name": "bash",3417                "permission_policy": {3418                  "type": "always_allow"3419                },3420                "type": "bash"3421              }3422            ],3423            "default_config": {3424              "enabled": true,3425              "permission_policy": {3426                "type": "always_ask"3427              }3428            },3429            "type": "agent_toolset_20260401"3430          }3431        ],3432        "type": "agent",3433        "version": 13434      },3435      "archived_at": null,3436      "budget": {3437        "max_list_cost": {3438          "amount": "2500",3439          "currency": "USD"3440        },3441        "type": "limit"3442      },3443      "created_at": "2026-03-15T10:00:00Z",3444      "environment_id": "env_011CZkZ9X2dpNyB7HsEFoRfW",3445      "metadata": {},3446      "outcome_evaluations": [3447        {3448          "completed_at": "2026-03-15T10:02:31Z",3449          "description": "Produce a 2-page summary as summary.md",3450          "explanation": "All five sections present with inline citations.",3451          "iteration": 0,3452          "outcome_id": "outc_011CZkZRSw2kEfs6ncTVljxP",3453          "result": "satisfied",3454          "type": "outcome_evaluation"3455        }3456      ],3457      "resources": [3458        {3459          "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",3460          "created_at": "2026-03-15T10:00:00Z",3461          "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",3462          "mount_path": "/uploads/receipt.pdf",3463          "type": "file",3464          "updated_at": "2026-03-15T10:00:00Z"3465        },3466        {3467          "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",3468          "created_at": "2026-03-15T10:00:00Z",3469          "mount_path": "/workspace/example-repo",3470          "type": "github_repository",3471          "updated_at": "2026-03-15T10:00:00Z",3472          "url": "https://github.com/example-org/example-repo",3473          "checkout": {3474            "name": "main",3475            "type": "branch"3476          }3477        }3478      ],3479      "stats": {3480        "active_seconds": 0,3481        "duration_seconds": 03482      },3483      "status": "idle",3484      "title": "Order #1234 inquiry",3485      "type": "session",3486      "updated_at": "2026-03-15T10:00:00Z",3487      "usage": {3488        "active_seconds": 0,3489        "cache_creation": {3490          "ephemeral_1h_input_tokens": 0,3491          "ephemeral_5m_input_tokens": 03492        },3493        "cache_read_input_tokens": 0,3494        "input_tokens": 0,3495        "list_cost": {3496          "amount": "2500",3497          "currency": "USD"3498        },3499        "output_tokens": 0,3500        "server_tool_use": {3501          "web_fetch_requests": 0,3502          "web_search_requests": 33503        }3504      },3505      "vault_ids": [3506        "vlt_011CZkZDLs7fYzm1hXNPeRjv"3507      ],3508      "deployment_id": "deployment_id"3509    }3510  ],3511  "next_page": "page_MjAyNS0wNS0xNFQwMDowMDowMFo=",3512  "prev_page": "page_MjAyNS0wNS0xM1QwMDowMDowMFo="3513}3514```35153516## Get Session35173518**GET** `/v1/sessions/{session_id}`35193520Get Session35213522### Path parameters35233524- `session_id: string`35253526### Headers35273528- `"anthropic-beta": optional array of AnthropicBeta`35293530  Optional header to specify the beta version(s) you want to use.35313532  - `string`35333534  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`35353536    - `"message-batches-2024-09-24"`35373538    - `"prompt-caching-2024-07-31"`35393540    - `"computer-use-2024-10-22"`35413542    - `"computer-use-2025-01-24"`35433544    - `"pdfs-2024-09-25"`35453546    - `"token-counting-2024-11-01"`35473548    - `"token-efficient-tools-2025-02-19"`35493550    - `"output-128k-2025-02-19"`35513552    - `"files-api-2025-04-14"`35533554    - `"mcp-client-2025-04-04"`35553556    - `"mcp-client-2025-11-20"`35573558    - `"dev-full-thinking-2025-05-14"`35593560    - `"interleaved-thinking-2025-05-14"`35613562    - `"code-execution-2025-05-22"`35633564    - `"extended-cache-ttl-2025-04-11"`35653566    - `"context-1m-2025-08-07"`35673568    - `"context-management-2025-06-27"`35693570    - `"model-context-window-exceeded-2025-08-26"`35713572    - `"skills-2025-10-02"`35733574    - `"fast-mode-2026-02-01"`35753576    - `"output-300k-2026-03-24"`35773578    - `"user-profiles-2026-03-24"`35793580    - `"user-profiles-2026-08-18"`35813582    - `"user-profiles-2026-09-04"`35833584    - `"advisor-tool-2026-03-01"`35853586    - `"managed-agents-2026-04-01"`35873588    - `"cache-diagnosis-2026-04-07"`35893590    - `"dreaming-2026-04-21"`35913592    - `"thinking-token-count-2026-05-13"`35933594    - `"server-side-fallback-2026-06-01"`35953596    - `"server-side-fallback-2026-07-01"`35973598    - `"fallback-credit-2026-06-01"`35993600    - `"fallback-credit-2026-07-01"`36013602    - `"agent-memory-2026-07-22"`36033604    - `"mid-conversation-tool-changes-2026-07-01"`36053606    - `"compact-2026-01-12"`36073608    - `"computer-use-2025-11-24"`36093610    - `"mcp-tunnels-2026-06-22"`36113612    - `"structured-outputs-2025-11-13"`36133614    - `"task-budgets-2026-03-13"`36153616    - `"thinking-display-updates-2026-08-18"`36173618    - `"ce-user-management-2026-07-13"`36193620    - `"mid-conversation-output-config-2026-07-01"`36213622    - `"thinking-binding-controls-2026-08-01"`36233624    - `"mid-conversation-system-clear-at-2026-08-21"`36253626    - `"compact-2026-09-04"`36273628- `"anthropic-workspace-id": optional string`36293630### Returns36313632- `BetaManagedAgentsSession object`36333634  A Managed Agents `session`.36353636  - `type: "session"`36373638  - `id: string`36393640  - `agent: BetaManagedAgentsSessionAgent`36413642    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.36433644    - `type: "agent"`36453646    - `id: string`36473648    - `description: string or null`36493650    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`36513652      - `type: "url"`36533654      - `name: string`36553656      - `url: string`36573658    - `model: BetaManagedAgentsModelConfig`36593660      Model identifier and configuration.36613662      - `id: BetaManagedAgentsModel`36633664        The model that will power your agent.36653666        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.36673668        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`36693670          The model that will power your agent.36713672          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.36733674          - `"claude-fable-5-1"`36753676            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows36773678          - `"claude-sonnet-5"`36793680            High-performance model for coding and agents36813682          - `"claude-fable-5"`36833684            Next generation of intelligence for the hardest knowledge work and coding problems36853686          - `"claude-opus-5"`36873688            Powerful intelligence for long-running agents and coding36893690          - `"claude-opus-4-8"`36913692            Powerful intelligence for long-running agents and coding36933694          - `"claude-opus-4-7"`36953696            Powerful intelligence for long-running agents and coding36973698          - `"claude-opus-4-6"`36993700            Powerful intelligence for long-running agents and coding37013702          - `"claude-sonnet-4-6"`37033704            Best combination of speed and intelligence37053706          - `"claude-haiku-4-5"`37073708            Fastest model with near-frontier intelligence37093710          - `"claude-haiku-4-5-20251001"`37113712            Fastest model with near-frontier intelligence37133714          - `"claude-opus-4-5"`37153716            Powerful intelligence for long-running agents and coding37173718          - `"claude-opus-4-5-20251101"`37193720            Powerful intelligence for long-running agents and coding37213722          - `"claude-sonnet-4-5"`37233724            High-performance model for agents and coding37253726          - `"claude-sonnet-4-5-20250929"`37273728            High-performance model for agents and coding37293730        - `string`37313732      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`37333734        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.37353736        - `BetaManagedAgentsEffortLow object`37373738          Low effort. Favors latency over reasoning depth.37393740          - `type: "low"`37413742        - `BetaManagedAgentsEffortMedium object`37433744          Medium effort. Balances latency and reasoning depth.37453746          - `type: "medium"`37473748        - `BetaManagedAgentsEffortHigh object`37493750          High effort. Favors reasoning depth.37513752          - `type: "high"`37533754        - `BetaManagedAgentsEffortXhigh object`37553756          Extra-high effort. Not all models accept this level.37573758          - `type: "xhigh"`37593760        - `BetaManagedAgentsEffortMax object`37613762          Maximum effort. Favors reasoning depth over latency.37633764          - `type: "max"`37653766      - `inference_geo: optional string`37673768        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.37693770      - `speed: optional "standard" or "fast"`37713772        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.37733774        - `"standard"`37753776        - `"fast"`37773778    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`37793780      Resolved coordinator topology with full agent definitions for each roster member.37813782      - `type: "coordinator"`37833784      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`37853786        Full `agent` definitions the coordinator may spawn as session threads.37873788        - `BetaManagedAgentsSessionThreadAgent object`37893790          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.37913792          - `type: "agent"`37933794          - `id: string`37953796          - `description: string or null`37973798          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`37993800            - `type: "url"`38013802            - `name: string`38033804            - `url: string`38053806          - `model: BetaManagedAgentsModelConfig`38073808            Model identifier and configuration.38093810          - `name: string`38113812          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`38133814            - `BetaManagedAgentsAnthropicSkill object`38153816              A resolved Anthropic-managed skill.38173818              - `type: "anthropic"`38193820              - `skill_id: string`38213822              - `version: string`38233824            - `BetaManagedAgentsCustomSkill object`38253826              A resolved user-created custom skill.38273828              - `type: "custom"`38293830              - `skill_id: string`38313832              - `version: string`38333834          - `system: string or null`38353836          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`38373838            - `BetaManagedAgentsAgentToolset20260401 object`38393840              - `type: "agent_toolset_20260401"`38413842              - `configs: array of BetaManagedAgentsAgentToolConfig`38433844                - `BetaManagedAgentsBashToolConfig object`38453846                  Configuration for the bash tool.38473848                  - `type: "bash"`38493850                  - `enabled: boolean`38513852                  - `name: "bash"`38533854                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`38553856                    Permission policy for tool execution.38573858                    - `BetaManagedAgentsAlwaysAllowPolicy object`38593860                      Tool calls are automatically approved without user confirmation.38613862                      - `type: "always_allow"`38633864                    - `BetaManagedAgentsAlwaysAskPolicy object`38653866                      Tool calls require user confirmation before execution.38673868                      - `type: "always_ask"`38693870                    - `BetaManagedAgentsAutoPolicy object`38713872                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.38733874                      - `type: "auto"`38753876                - `BetaManagedAgentsEditToolConfig object`38773878                  Configuration for the edit tool.38793880                  - `type: "edit"`38813882                  - `enabled: boolean`38833884                  - `name: "edit"`38853886                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`38873888                    Permission policy for tool execution.38893890                    - `BetaManagedAgentsAlwaysAllowPolicy object`38913892                      Tool calls are automatically approved without user confirmation.38933894                    - `BetaManagedAgentsAlwaysAskPolicy object`38953896                      Tool calls require user confirmation before execution.38973898                    - `BetaManagedAgentsAutoPolicy object`38993900                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.39013902                - `BetaManagedAgentsReadToolConfig object`39033904                  Configuration for the read tool.39053906                  - `type: "read"`39073908                  - `enabled: boolean`39093910                  - `name: "read"`39113912                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`39133914                    Permission policy for tool execution.39153916                    - `BetaManagedAgentsAlwaysAllowPolicy object`39173918                      Tool calls are automatically approved without user confirmation.39193920                    - `BetaManagedAgentsAlwaysAskPolicy object`39213922                      Tool calls require user confirmation before execution.39233924                    - `BetaManagedAgentsAutoPolicy object`39253926                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.39273928                - `BetaManagedAgentsWriteToolConfig object`39293930                  Configuration for the write tool.39313932                  - `type: "write"`39333934                  - `enabled: boolean`39353936                  - `name: "write"`39373938                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`39393940                    Permission policy for tool execution.39413942                    - `BetaManagedAgentsAlwaysAllowPolicy object`39433944                      Tool calls are automatically approved without user confirmation.39453946                    - `BetaManagedAgentsAlwaysAskPolicy object`39473948                      Tool calls require user confirmation before execution.39493950                    - `BetaManagedAgentsAutoPolicy object`39513952                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.39533954                - `BetaManagedAgentsGlobToolConfig object`39553956                  Configuration for the glob tool.39573958                  - `type: "glob"`39593960                  - `enabled: boolean`39613962                  - `name: "glob"`39633964                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`39653966                    Permission policy for tool execution.39673968                    - `BetaManagedAgentsAlwaysAllowPolicy object`39693970                      Tool calls are automatically approved without user confirmation.39713972                    - `BetaManagedAgentsAlwaysAskPolicy object`39733974                      Tool calls require user confirmation before execution.39753976                    - `BetaManagedAgentsAutoPolicy object`39773978                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.39793980                - `BetaManagedAgentsGrepToolConfig object`39813982                  Configuration for the grep tool.39833984                  - `type: "grep"`39853986                  - `enabled: boolean`39873988                  - `name: "grep"`39893990                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`39913992                    Permission policy for tool execution.39933994                    - `BetaManagedAgentsAlwaysAllowPolicy object`39953996                      Tool calls are automatically approved without user confirmation.39973998                    - `BetaManagedAgentsAlwaysAskPolicy object`39994000                      Tool calls require user confirmation before execution.40014002                    - `BetaManagedAgentsAutoPolicy object`40034004                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.40054006                - `BetaManagedAgentsWebFetchToolConfig object`40074008                  Configuration for the web_fetch tool.40094010                  - `type: "web_fetch"`40114012                  - `enabled: boolean`40134014                  - `name: "web_fetch"`40154016                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`40174018                    Permission policy for tool execution.40194020                    - `BetaManagedAgentsAlwaysAllowPolicy object`40214022                      Tool calls are automatically approved without user confirmation.40234024                    - `BetaManagedAgentsAlwaysAskPolicy object`40254026                      Tool calls require user confirmation before execution.40274028                    - `BetaManagedAgentsAutoPolicy object`40294030                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.40314032                  - `allowed_domains: optional array of string`40334034                  - `blocked_domains: optional array of string`40354036                  - `max_content_tokens: optional number or null`40374038                    format: int3240394040                - `BetaManagedAgentsWebSearchToolConfig object`40414042                  Configuration for the web_search tool.40434044                  - `type: "web_search"`40454046                  - `enabled: boolean`40474048                  - `name: "web_search"`40494050                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`40514052                    Permission policy for tool execution.40534054                    - `BetaManagedAgentsAlwaysAllowPolicy object`40554056                      Tool calls are automatically approved without user confirmation.40574058                    - `BetaManagedAgentsAlwaysAskPolicy object`40594060                      Tool calls require user confirmation before execution.40614062                    - `BetaManagedAgentsAutoPolicy object`40634064                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.40654066                  - `allowed_domains: optional array of string`40674068                  - `blocked_domains: optional array of string`40694070                  - `user_location: optional BetaManagedAgentsUserLocation or null`40714072                    Approximate user location for search result localization.40734074                    - `type: "approximate"`40754076                      Location precision. Only "approximate" is supported.40774078                    - `city: optional string or null`40794080                      City name.40814082                      minLength: 1, maxLength: 25540834084                    - `country: optional string or null`40854086                      Two-letter ISO 3166-1 country code, uppercase.40874088                    - `region: optional string or null`40894090                      Region or state name.40914092                      minLength: 1, maxLength: 25540934094                    - `timezone: optional string or null`40954096                      IANA timezone identifier, e.g. "America/Los_Angeles".40974098                      minLength: 1, maxLength: 25540994100              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`41014102                Resolved default configuration for agent tools.41034104                - `enabled: boolean`41054106                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`41074108                  Permission policy for tool execution.41094110                  - `BetaManagedAgentsAlwaysAllowPolicy object`41114112                    Tool calls are automatically approved without user confirmation.41134114                  - `BetaManagedAgentsAlwaysAskPolicy object`41154116                    Tool calls require user confirmation before execution.41174118                  - `BetaManagedAgentsAutoPolicy object`41194120                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.41214122            - `BetaManagedAgentsMCPToolset object`41234124              - `type: "mcp_toolset"`41254126              - `configs: array of BetaManagedAgentsMCPToolConfig`41274128                - `enabled: boolean`41294130                - `name: string`41314132                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`41334134                  Permission policy for tool execution.41354136                  - `BetaManagedAgentsAlwaysAllowPolicy object`41374138                    Tool calls are automatically approved without user confirmation.41394140                  - `BetaManagedAgentsAlwaysAskPolicy object`41414142                    Tool calls require user confirmation before execution.41434144                  - `BetaManagedAgentsAutoPolicy object`41454146                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.41474148              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`41494150                Resolved default configuration for all tools from an MCP server.41514152                - `enabled: boolean`41534154                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`41554156                  Permission policy for tool execution.41574158                  - `BetaManagedAgentsAlwaysAllowPolicy object`41594160                    Tool calls are automatically approved without user confirmation.41614162                  - `BetaManagedAgentsAlwaysAskPolicy object`41634164                    Tool calls require user confirmation before execution.41654166                  - `BetaManagedAgentsAutoPolicy object`41674168                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.41694170              - `mcp_server_name: string`41714172            - `BetaManagedAgentsCustomTool object`41734174              A custom tool as returned in API responses.41754176              - `type: "custom"`41774178              - `description: string`41794180              - `input_schema: BetaManagedAgentsCustomToolInputSchema`41814182                JSON Schema for custom tool input parameters.41834184                - `type: "object"`41854186                - `properties: optional map[unknown] or null`41874188                - `required: optional array of string or null`41894190              - `name: string`41914192          - `version: number`41934194            format: int3241954196        - `BetaManagedAgentsAdvisor object`41974198          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.41994200          - `type: "advisor"`42014202          - `model: string`42034204            The advisor model id.42054206    - `name: string`42074208    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`42094210      - `BetaManagedAgentsAnthropicSkill object`42114212        A resolved Anthropic-managed skill.42134214      - `BetaManagedAgentsCustomSkill object`42154216        A resolved user-created custom skill.42174218    - `system: string or null`42194220    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`42214222      - `BetaManagedAgentsAgentToolset20260401 object`42234224      - `BetaManagedAgentsMCPToolset object`42254226      - `BetaManagedAgentsCustomTool object`42274228        A custom tool as returned in API responses.42294230    - `version: number`42314232      format: int3242334234  - `archived_at: string or null`42354236    A timestamp in RFC 3339 format42374238    format: date-time42394240  - `budget: BetaManagedAgentsBudgetLimit or null`42414242    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.42434244    - `type: "limit"`42454246    - `max_list_cost: BetaMonetaryAmount`42474248      A monetary amount in a specific currency.42494250      - `amount: string`42514252        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.42534254      - `currency: BetaCurrency`42554256        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.42574258  - `created_at: string`42594260    A timestamp in RFC 3339 format42614262    format: date-time42634264  - `environment_id: string`42654266  - `metadata: map[string]`42674268  - `outcome_evaluations: array of BetaManagedAgentsOutcomeEvaluationResource`42694270    Per-outcome evaluation state. One entry per `define_outcome` event sent to the session.42714272    - `type: "outcome_evaluation"`42734274    - `completed_at: string or null`42754276      A timestamp in RFC 3339 format42774278      format: date-time42794280    - `description: string`42814282      What the agent should produce.42834284    - `explanation: string or null`42854286      Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.42874288    - `iteration: number`42894290      0-indexed revision cycle the outcome is currently on.42914292      format: int3242934294    - `outcome_id: string`42954296      Server-generated outc_ ID for this outcome.42974298    - `result: string`42994300      Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.43014302  - `resources: array of BetaManagedAgentsSessionResource`43034304    - `BetaManagedAgentsGitHubRepositoryResource object`43054306      - `type: "github_repository"`43074308      - `id: string`43094310      - `created_at: string`43114312        A timestamp in RFC 3339 format43134314        format: date-time43154316      - `mount_path: string`43174318      - `updated_at: string`43194320        A timestamp in RFC 3339 format43214322        format: date-time43234324      - `url: string`43254326      - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`43274328        - `BetaManagedAgentsBranchCheckout object`43294330          - `type: "branch"`43314332          - `name: string`43334334            Branch name to check out.43354336            minLength: 1, maxLength: 25543374338        - `BetaManagedAgentsCommitCheckout object`43394340          - `type: "commit"`43414342          - `sha: string`43434344            Full commit SHA to check out.43454346            minLength: 7, maxLength: 6443474348    - `BetaManagedAgentsFileResource object`43494350      - `type: "file"`43514352      - `id: string`43534354      - `created_at: string`43554356        A timestamp in RFC 3339 format43574358        format: date-time43594360      - `file_id: string`43614362      - `mount_path: string`43634364      - `updated_at: string`43654366        A timestamp in RFC 3339 format43674368        format: date-time43694370    - `BetaManagedAgentsMemoryStoreResource object`43714372      A memory store attached to an agent session.43734374      - `type: "memory_store"`43754376      - `memory_store_id: string`43774378        The memory store ID (memstore_...). Must belong to the caller's organization and workspace.43794380      - `access: optional "read_write" or "read_only" or null`43814382        Access mode for an attached memory store.43834384        - `"read_write"`43854386        - `"read_only"`43874388      - `description: optional string`43894390        Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.43914392      - `instructions: optional string or null`43934394        Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.43954396        maxLength: 409643974398      - `mount_path: optional string or null`43994400        Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.44014402      - `name: optional string or null`44034404        Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.44054406  - `stats: BetaManagedAgentsSessionStats`44074408    Timing statistics for a session.44094410    - `active_seconds: optional number`44114412      Cumulative time in seconds the session spent in `running` status. Excludes idle time.44134414      format: double44154416    - `duration_seconds: optional number`44174418      Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.44194420      format: double44214422  - `status: "rescheduling" or "running" or "idle" or "terminated"`44234424    SessionStatus enum44254426    - `"rescheduling"`44274428    - `"running"`44294430    - `"idle"`44314432    - `"terminated"`44334434  - `title: string or null`44354436  - `updated_at: string`44374438    A timestamp in RFC 3339 format44394440    format: date-time44414442  - `usage: BetaManagedAgentsSessionUsage`44434444    Cumulative token usage for a session across all turns.44454446    - `active_seconds: optional number`44474448      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.44494450      format: double44514452    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`44534454      Prompt-cache creation token usage broken down by cache lifetime.44554456      - `ephemeral_1h_input_tokens: optional number`44574458        Tokens used to create 1-hour ephemeral cache entries.44594460        format: int3244614462      - `ephemeral_5m_input_tokens: optional number`44634464        Tokens used to create 5-minute ephemeral cache entries.44654466        format: int3244674468    - `cache_read_input_tokens: optional number`44694470      Total tokens read from prompt cache.44714472      format: int3244734474    - `input_tokens: optional number`44754476      Total input tokens consumed across all turns.44774478      format: int3244794480    - `list_cost: optional BetaMonetaryAmount or null`44814482      A monetary amount in a specific currency.44834484    - `output_tokens: optional number`44854486      Total output tokens generated across all turns.44874488      format: int3244894490    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`44914492      Cumulative count of server-executed tool invocations, broken down by tool.44934494      - `web_fetch_requests: optional number`44954496        Number of server-executed web fetch requests.44974498        format: int3244994500      - `web_search_requests: optional number`45014502        Number of server-executed web search requests.45034504        format: int3245054506  - `vault_ids: array of string`45074508    Vault IDs attached to the session at creation. Empty when no vaults were supplied.45094510  - `deployment_id: optional string or null`45114512    Deployment ID when the session was created from a deployment reference. Null otherwise.45134514### Example45154516```bash4517curl https://api.anthropic.com/v1/sessions/$SESSION_ID \4518    -H 'anthropic-version: 2023-06-01' \4519    -H 'anthropic-beta: managed-agents-2026-04-01' \4520    -H "X-Api-Key: $ANTHROPIC_API_KEY"4521```45224523#### Response (200)45244525```json4526{4527  "id": "sesn_011CZkZAtmR3yMPDzynEDxu7",4528  "agent": {4529    "id": "agent_011CZkYpogX7uDKUyvBTophP",4530    "description": "A general-purpose starter agent.",4531    "mcp_servers": [4532      {4533        "name": "example-mcp",4534        "type": "url",4535        "url": "https://example-server.modelcontextprotocol.io/sse"4536      }4537    ],4538    "model": {4539      "id": "claude-opus-5",4540      "effort": {4541        "type": "low"4542      },4543      "inference_geo": "inference_geo",4544      "speed": "standard"4545    },4546    "multiagent": {4547      "agents": [4548        {4549          "id": "agent_011CZkYqphY8vELVzwCUpqiQ",4550          "description": "A focused research subagent.",4551          "mcp_servers": [4552            {4553              "name": "example-mcp",4554              "type": "url",4555              "url": "https://example-server.modelcontextprotocol.io/sse"4556            }4557          ],4558          "model": {4559            "id": "claude-opus-5",4560            "effort": {4561              "type": "low"4562            },4563            "inference_geo": "inference_geo",4564            "speed": "standard"4565          },4566          "name": "Researcher",4567          "skills": [4568            {4569              "skill_id": "xlsx",4570              "type": "anthropic",4571              "version": "1"4572            }4573          ],4574          "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",4575          "tools": [4576            {4577              "configs": [4578                {4579                  "enabled": true,4580                  "name": "bash",4581                  "permission_policy": {4582                    "type": "always_allow"4583                  },4584                  "type": "bash"4585                }4586              ],4587              "default_config": {4588                "enabled": true,4589                "permission_policy": {4590                  "type": "always_ask"4591                }4592              },4593              "type": "agent_toolset_20260401"4594            }4595          ],4596          "type": "agent",4597          "version": 14598        }4599      ],4600      "type": "coordinator"4601    },4602    "name": "My First Agent",4603    "skills": [4604      {4605        "skill_id": "xlsx",4606        "type": "anthropic",4607        "version": "1"4608      },4609      {4610        "skill_id": "skill_011CZkZFNu9hAbo3jZPRgTlx",4611        "type": "custom",4612        "version": "2"4613      }4614    ],4615    "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user's task end to end.",4616    "tools": [4617      {4618        "configs": [4619          {4620            "enabled": true,4621            "name": "bash",4622            "permission_policy": {4623              "type": "always_allow"4624            },4625            "type": "bash"4626          }4627        ],4628        "default_config": {4629          "enabled": true,4630          "permission_policy": {4631            "type": "always_ask"4632          }4633        },4634        "type": "agent_toolset_20260401"4635      }4636    ],4637    "type": "agent",4638    "version": 14639  },4640  "archived_at": null,4641  "budget": {4642    "max_list_cost": {4643      "amount": "2500",4644      "currency": "USD"4645    },4646    "type": "limit"4647  },4648  "created_at": "2026-03-15T10:00:00Z",4649  "environment_id": "env_011CZkZ9X2dpNyB7HsEFoRfW",4650  "metadata": {},4651  "outcome_evaluations": [4652    {4653      "completed_at": "2026-03-15T10:02:31Z",4654      "description": "Produce a 2-page summary as summary.md",4655      "explanation": "All five sections present with inline citations.",4656      "iteration": 0,4657      "outcome_id": "outc_011CZkZRSw2kEfs6ncTVljxP",4658      "result": "satisfied",4659      "type": "outcome_evaluation"4660    }4661  ],4662  "resources": [4663    {4664      "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",4665      "created_at": "2026-03-15T10:00:00Z",4666      "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",4667      "mount_path": "/uploads/receipt.pdf",4668      "type": "file",4669      "updated_at": "2026-03-15T10:00:00Z"4670    },4671    {4672      "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",4673      "created_at": "2026-03-15T10:00:00Z",4674      "mount_path": "/workspace/example-repo",4675      "type": "github_repository",4676      "updated_at": "2026-03-15T10:00:00Z",4677      "url": "https://github.com/example-org/example-repo",4678      "checkout": {4679        "name": "main",4680        "type": "branch"4681      }4682    }4683  ],4684  "stats": {4685    "active_seconds": 0,4686    "duration_seconds": 04687  },4688  "status": "idle",4689  "title": "Order #1234 inquiry",4690  "type": "session",4691  "updated_at": "2026-03-15T10:00:00Z",4692  "usage": {4693    "active_seconds": 0,4694    "cache_creation": {4695      "ephemeral_1h_input_tokens": 0,4696      "ephemeral_5m_input_tokens": 04697    },4698    "cache_read_input_tokens": 0,4699    "input_tokens": 0,4700    "list_cost": {4701      "amount": "2500",4702      "currency": "USD"4703    },4704    "output_tokens": 0,4705    "server_tool_use": {4706      "web_fetch_requests": 0,4707      "web_search_requests": 34708    }4709  },4710  "vault_ids": [4711    "vlt_011CZkZDLs7fYzm1hXNPeRjv"4712  ],4713  "deployment_id": "deployment_id"4714}4715```47164717## Update Session47184719**POST** `/v1/sessions/{session_id}`47204721Update Session47224723### Path parameters47244725- `session_id: string`47264727### Headers47284729- `"anthropic-beta": optional array of AnthropicBeta`47304731  Optional header to specify the beta version(s) you want to use.47324733  - `string`47344735  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`47364737    - `"message-batches-2024-09-24"`47384739    - `"prompt-caching-2024-07-31"`47404741    - `"computer-use-2024-10-22"`47424743    - `"computer-use-2025-01-24"`47444745    - `"pdfs-2024-09-25"`47464747    - `"token-counting-2024-11-01"`47484749    - `"token-efficient-tools-2025-02-19"`47504751    - `"output-128k-2025-02-19"`47524753    - `"files-api-2025-04-14"`47544755    - `"mcp-client-2025-04-04"`47564757    - `"mcp-client-2025-11-20"`47584759    - `"dev-full-thinking-2025-05-14"`47604761    - `"interleaved-thinking-2025-05-14"`47624763    - `"code-execution-2025-05-22"`47644765    - `"extended-cache-ttl-2025-04-11"`47664767    - `"context-1m-2025-08-07"`47684769    - `"context-management-2025-06-27"`47704771    - `"model-context-window-exceeded-2025-08-26"`47724773    - `"skills-2025-10-02"`47744775    - `"fast-mode-2026-02-01"`47764777    - `"output-300k-2026-03-24"`47784779    - `"user-profiles-2026-03-24"`47804781    - `"user-profiles-2026-08-18"`47824783    - `"user-profiles-2026-09-04"`47844785    - `"advisor-tool-2026-03-01"`47864787    - `"managed-agents-2026-04-01"`47884789    - `"cache-diagnosis-2026-04-07"`47904791    - `"dreaming-2026-04-21"`47924793    - `"thinking-token-count-2026-05-13"`47944795    - `"server-side-fallback-2026-06-01"`47964797    - `"server-side-fallback-2026-07-01"`47984799    - `"fallback-credit-2026-06-01"`48004801    - `"fallback-credit-2026-07-01"`48024803    - `"agent-memory-2026-07-22"`48044805    - `"mid-conversation-tool-changes-2026-07-01"`48064807    - `"compact-2026-01-12"`48084809    - `"computer-use-2025-11-24"`48104811    - `"mcp-tunnels-2026-06-22"`48124813    - `"structured-outputs-2025-11-13"`48144815    - `"task-budgets-2026-03-13"`48164817    - `"thinking-display-updates-2026-08-18"`48184819    - `"ce-user-management-2026-07-13"`48204821    - `"mid-conversation-output-config-2026-07-01"`48224823    - `"thinking-binding-controls-2026-08-01"`48244825    - `"mid-conversation-system-clear-at-2026-08-21"`48264827    - `"compact-2026-09-04"`48284829- `"anthropic-workspace-id": optional string`48304831### Body parameters48324833- `agent: optional BetaManagedAgentsSessionAgentUpdate`48344835  Mid-session agent configuration update. Only `tools` and `mcp_servers` are updatable. Full replacement: the provided array becomes the new value. To preserve existing entries, GET the session, modify the array, and POST it back.48364837  - `mcp_servers: optional array of BetaManagedAgentsURLMCPServerParams`48384839    Replacement MCP server list. Full replacement: the provided array becomes the new value. Send an empty array to clear; omit to preserve.48404841    - `type: "url"`48424843    - `name: string`48444845      Unique name for this server, referenced by mcp_toolset configurations. 1-255 characters.48464847      minLength: 1, maxLength: 25548484849    - `url: string`48504851      Endpoint URL for the MCP server.48524853      maxLength: 204848544855  - `tools: optional array of BetaManagedAgentsAgentToolset20260401Params or BetaManagedAgentsMCPToolsetParams or BetaManagedAgentsCustomToolParams`48564857    Replacement tool list. Full replacement: the provided array becomes the new value. Send an empty array to clear; omit to preserve.48584859    - `BetaManagedAgentsAgentToolset20260401Params object`48604861      Configuration for built-in agent tools. Use this to enable or disable groups of tools available to the agent.48624863      - `type: "agent_toolset_20260401"`48644865      - `configs: optional array of BetaManagedAgentsAgentToolConfigParams`48664867        Per-tool configuration overrides.48684869        - `BetaManagedAgentsBashToolConfigParams object`48704871          Configuration override for the bash tool.48724873          - `type: optional "bash"`48744875          - `name: "bash"`48764877            Must be "bash".48784879          - `enabled: optional boolean or null`48804881            Whether this tool is enabled and available to Claude. Overrides the default_config setting.48824883          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`48844885            Permission policy for tool execution.48864887            - `BetaManagedAgentsAlwaysAllowPolicy object`48884889              Tool calls are automatically approved without user confirmation.48904891              - `type: "always_allow"`48924893            - `BetaManagedAgentsAlwaysAskPolicy object`48944895              Tool calls require user confirmation before execution.48964897              - `type: "always_ask"`48984899            - `BetaManagedAgentsAutoPolicy object`49004901              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.49024903              - `type: "auto"`49044905        - `BetaManagedAgentsEditToolConfigParams object`49064907          Configuration override for the edit tool.49084909          - `type: optional "edit"`49104911          - `name: "edit"`49124913            Must be "edit".49144915          - `enabled: optional boolean or null`49164917            Whether this tool is enabled and available to Claude. Overrides the default_config setting.49184919          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`49204921            Permission policy for tool execution.49224923            - `BetaManagedAgentsAlwaysAllowPolicy object`49244925              Tool calls are automatically approved without user confirmation.49264927            - `BetaManagedAgentsAlwaysAskPolicy object`49284929              Tool calls require user confirmation before execution.49304931            - `BetaManagedAgentsAutoPolicy object`49324933              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.49344935        - `BetaManagedAgentsReadToolConfigParams object`49364937          Configuration override for the read tool.49384939          - `type: optional "read"`49404941          - `name: "read"`49424943            Must be "read".49444945          - `enabled: optional boolean or null`49464947            Whether this tool is enabled and available to Claude. Overrides the default_config setting.49484949          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`49504951            Permission policy for tool execution.49524953            - `BetaManagedAgentsAlwaysAllowPolicy object`49544955              Tool calls are automatically approved without user confirmation.49564957            - `BetaManagedAgentsAlwaysAskPolicy object`49584959              Tool calls require user confirmation before execution.49604961            - `BetaManagedAgentsAutoPolicy object`49624963              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.49644965        - `BetaManagedAgentsWriteToolConfigParams object`49664967          Configuration override for the write tool.49684969          - `type: optional "write"`49704971          - `name: "write"`49724973            Must be "write".49744975          - `enabled: optional boolean or null`49764977            Whether this tool is enabled and available to Claude. Overrides the default_config setting.49784979          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`49804981            Permission policy for tool execution.49824983            - `BetaManagedAgentsAlwaysAllowPolicy object`49844985              Tool calls are automatically approved without user confirmation.49864987            - `BetaManagedAgentsAlwaysAskPolicy object`49884989              Tool calls require user confirmation before execution.49904991            - `BetaManagedAgentsAutoPolicy object`49924993              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.49944995        - `BetaManagedAgentsGlobToolConfigParams object`49964997          Configuration override for the glob tool.49984999          - `type: optional "glob"`50005001          - `name: "glob"`50025003            Must be "glob".50045005          - `enabled: optional boolean or null`50065007            Whether this tool is enabled and available to Claude. Overrides the default_config setting.50085009          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`50105011            Permission policy for tool execution.50125013            - `BetaManagedAgentsAlwaysAllowPolicy object`50145015              Tool calls are automatically approved without user confirmation.50165017            - `BetaManagedAgentsAlwaysAskPolicy object`50185019              Tool calls require user confirmation before execution.50205021            - `BetaManagedAgentsAutoPolicy object`50225023              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.50245025        - `BetaManagedAgentsGrepToolConfigParams object`50265027          Configuration override for the grep tool.50285029          - `type: optional "grep"`50305031          - `name: "grep"`50325033            Must be "grep".50345035          - `enabled: optional boolean or null`50365037            Whether this tool is enabled and available to Claude. Overrides the default_config setting.50385039          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`50405041            Permission policy for tool execution.50425043            - `BetaManagedAgentsAlwaysAllowPolicy object`50445045              Tool calls are automatically approved without user confirmation.50465047            - `BetaManagedAgentsAlwaysAskPolicy object`50485049              Tool calls require user confirmation before execution.50505051            - `BetaManagedAgentsAutoPolicy object`50525053              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.50545055        - `BetaManagedAgentsWebFetchToolConfigParams object`50565057          Configuration override for the web_fetch tool.50585059          - `type: optional "web_fetch"`50605061          - `name: "web_fetch"`50625063            Must be "web_fetch".50645065          - `allowed_domains: optional array of string`50665067            Only fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.50685069          - `blocked_domains: optional array of string`50705071            Never fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.50725073          - `enabled: optional boolean or null`50745075            Whether this tool is enabled and available to Claude. Overrides the default_config setting.50765077          - `max_content_tokens: optional number or null`50785079            Maximum number of tokens of fetched text content to include in context per call. Does not apply to binary content such as PDFs.50805081            format: int3250825083          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`50845085            Permission policy for tool execution.50865087            - `BetaManagedAgentsAlwaysAllowPolicy object`50885089              Tool calls are automatically approved without user confirmation.50905091            - `BetaManagedAgentsAlwaysAskPolicy object`50925093              Tool calls require user confirmation before execution.50945095            - `BetaManagedAgentsAutoPolicy object`50965097              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.50985099        - `BetaManagedAgentsWebSearchToolConfigParams object`51005101          Configuration override for the web_search tool.51025103          - `type: optional "web_search"`51045105          - `name: "web_search"`51065107            Must be "web_search".51085109          - `allowed_domains: optional array of string`51105111            Only return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.51125113          - `blocked_domains: optional array of string`51145115            Never return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.51165117          - `enabled: optional boolean or null`51185119            Whether this tool is enabled and available to Claude. Overrides the default_config setting.51205121          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`51225123            Permission policy for tool execution.51245125            - `BetaManagedAgentsAlwaysAllowPolicy object`51265127              Tool calls are automatically approved without user confirmation.51285129            - `BetaManagedAgentsAlwaysAskPolicy object`51305131              Tool calls require user confirmation before execution.51325133            - `BetaManagedAgentsAutoPolicy object`51345135              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.51365137          - `user_location: optional BetaManagedAgentsUserLocation or null`51385139            Approximate user location for search result localization.51405141            - `type: "approximate"`51425143              Location precision. Only "approximate" is supported.51445145            - `city: optional string or null`51465147              City name.51485149              minLength: 1, maxLength: 25551505151            - `country: optional string or null`51525153              Two-letter ISO 3166-1 country code, uppercase.51545155            - `region: optional string or null`51565157              Region or state name.51585159              minLength: 1, maxLength: 25551605161            - `timezone: optional string or null`51625163              IANA timezone identifier, e.g. "America/Los_Angeles".51645165              minLength: 1, maxLength: 25551665167      - `default_config: optional BetaManagedAgentsAgentToolsetDefaultConfigParams or null`51685169        Default configuration for all tools in a toolset.51705171        - `enabled: optional boolean or null`51725173          Whether tools are enabled and available to Claude by default. Defaults to true if not specified.51745175        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`51765177          Permission policy for tool execution.51785179          - `BetaManagedAgentsAlwaysAllowPolicy object`51805181            Tool calls are automatically approved without user confirmation.51825183          - `BetaManagedAgentsAlwaysAskPolicy object`51845185            Tool calls require user confirmation before execution.51865187          - `BetaManagedAgentsAutoPolicy object`51885189            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.51905191    - `BetaManagedAgentsMCPToolsetParams object`51925193      Configuration for tools from an MCP server defined in `mcp_servers`.51945195      - `type: "mcp_toolset"`51965197      - `mcp_server_name: string`51985199        Name of the MCP server. Must match a server name from the mcp_servers array. 1-255 characters.52005201        minLength: 1, maxLength: 25552025203      - `configs: optional array of BetaManagedAgentsMCPToolConfigParams`52045205        Per-tool configuration overrides.52065207        - `name: string`52085209          Name of the MCP tool to configure. 1-128 characters.52105211          minLength: 1, maxLength: 12852125213        - `enabled: optional boolean or null`52145215          Whether this tool is enabled. Overrides the `default_config` setting.52165217        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`52185219          Permission policy for tool execution.52205221          - `BetaManagedAgentsAlwaysAllowPolicy object`52225223            Tool calls are automatically approved without user confirmation.52245225          - `BetaManagedAgentsAlwaysAskPolicy object`52265227            Tool calls require user confirmation before execution.52285229          - `BetaManagedAgentsAutoPolicy object`52305231            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.52325233      - `default_config: optional BetaManagedAgentsMCPToolsetDefaultConfigParams or null`52345235        Default configuration for all tools from an MCP server.52365237        - `enabled: optional boolean or null`52385239          Whether tools are enabled by default. Defaults to true if not specified.52405241        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`52425243          Permission policy for tool execution.52445245          - `BetaManagedAgentsAlwaysAllowPolicy object`52465247            Tool calls are automatically approved without user confirmation.52485249          - `BetaManagedAgentsAlwaysAskPolicy object`52505251            Tool calls require user confirmation before execution.52525253          - `BetaManagedAgentsAutoPolicy object`52545255            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.52565257    - `BetaManagedAgentsCustomToolParams object`52585259      A custom tool that is executed by the API client rather than the agent. When the agent calls this tool, an `agent.custom_tool_use` event is emitted and the session goes idle, waiting for the client to provide the result via a `user.custom_tool_result` event.52605261      - `type: "custom"`52625263      - `description: string`52645265        Description of what the tool does, shown to the agent to help it decide when to use the tool.52665267        minLength: 152685269      - `input_schema: BetaManagedAgentsCustomToolInputSchema`52705271        JSON Schema for custom tool input parameters.52725273        - `type: "object"`52745275        - `properties: optional map[unknown] or null`52765277        - `required: optional array of string or null`52785279      - `name: string`52805281        Unique name for the tool. 1-128 characters; letters, digits, underscores, and hyphens.52825283        minLength: 1, maxLength: 12852845285- `budget: optional BetaManagedAgentsBudgetLimit or null`52865287  A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.52885289  - `type: "limit"`52905291  - `max_list_cost: BetaMonetaryAmount`52925293    A monetary amount in a specific currency.52945295    - `amount: string`52965297      Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.52985299    - `currency: BetaCurrency`53005301      Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.53025303- `metadata: optional map[string] or null`53045305  Metadata patch. Set a key to a string to upsert it, or to null to delete it. Omit the field to preserve.53065307- `title: optional string or null`53085309  Human-readable session title.53105311  minLength: 1, maxLength: 50053125313- `vault_ids: optional array of string`53145315  Vault IDs (`vlt_*`) to attach to the session. Not yet supported; requests setting this field are rejected. Reserved for future use.53165317### Returns53185319- `BetaManagedAgentsSession object`53205321  A Managed Agents `session`.53225323  - `type: "session"`53245325  - `id: string`53265327  - `agent: BetaManagedAgentsSessionAgent`53285329    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.53305331    - `type: "agent"`53325333    - `id: string`53345335    - `description: string or null`53365337    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`53385339      - `type: "url"`53405341      - `name: string`53425343      - `url: string`53445345    - `model: BetaManagedAgentsModelConfig`53465347      Model identifier and configuration.53485349      - `id: BetaManagedAgentsModel`53505351        The model that will power your agent.53525353        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.53545355        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`53565357          The model that will power your agent.53585359          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.53605361          - `"claude-fable-5-1"`53625363            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows53645365          - `"claude-sonnet-5"`53665367            High-performance model for coding and agents53685369          - `"claude-fable-5"`53705371            Next generation of intelligence for the hardest knowledge work and coding problems53725373          - `"claude-opus-5"`53745375            Powerful intelligence for long-running agents and coding53765377          - `"claude-opus-4-8"`53785379            Powerful intelligence for long-running agents and coding53805381          - `"claude-opus-4-7"`53825383            Powerful intelligence for long-running agents and coding53845385          - `"claude-opus-4-6"`53865387            Powerful intelligence for long-running agents and coding53885389          - `"claude-sonnet-4-6"`53905391            Best combination of speed and intelligence53925393          - `"claude-haiku-4-5"`53945395            Fastest model with near-frontier intelligence53965397          - `"claude-haiku-4-5-20251001"`53985399            Fastest model with near-frontier intelligence54005401          - `"claude-opus-4-5"`54025403            Powerful intelligence for long-running agents and coding54045405          - `"claude-opus-4-5-20251101"`54065407            Powerful intelligence for long-running agents and coding54085409          - `"claude-sonnet-4-5"`54105411            High-performance model for agents and coding54125413          - `"claude-sonnet-4-5-20250929"`54145415            High-performance model for agents and coding54165417        - `string`54185419      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`54205421        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.54225423        - `BetaManagedAgentsEffortLow object`54245425          Low effort. Favors latency over reasoning depth.54265427          - `type: "low"`54285429        - `BetaManagedAgentsEffortMedium object`54305431          Medium effort. Balances latency and reasoning depth.54325433          - `type: "medium"`54345435        - `BetaManagedAgentsEffortHigh object`54365437          High effort. Favors reasoning depth.54385439          - `type: "high"`54405441        - `BetaManagedAgentsEffortXhigh object`54425443          Extra-high effort. Not all models accept this level.54445445          - `type: "xhigh"`54465447        - `BetaManagedAgentsEffortMax object`54485449          Maximum effort. Favors reasoning depth over latency.54505451          - `type: "max"`54525453      - `inference_geo: optional string`54545455        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.54565457      - `speed: optional "standard" or "fast"`54585459        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.54605461        - `"standard"`54625463        - `"fast"`54645465    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`54665467      Resolved coordinator topology with full agent definitions for each roster member.54685469      - `type: "coordinator"`54705471      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`54725473        Full `agent` definitions the coordinator may spawn as session threads.54745475        - `BetaManagedAgentsSessionThreadAgent object`54765477          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.54785479          - `type: "agent"`54805481          - `id: string`54825483          - `description: string or null`54845485          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`54865487            - `type: "url"`54885489            - `name: string`54905491            - `url: string`54925493          - `model: BetaManagedAgentsModelConfig`54945495            Model identifier and configuration.54965497          - `name: string`54985499          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`55005501            - `BetaManagedAgentsAnthropicSkill object`55025503              A resolved Anthropic-managed skill.55045505              - `type: "anthropic"`55065507              - `skill_id: string`55085509              - `version: string`55105511            - `BetaManagedAgentsCustomSkill object`55125513              A resolved user-created custom skill.55145515              - `type: "custom"`55165517              - `skill_id: string`55185519              - `version: string`55205521          - `system: string or null`55225523          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`55245525            - `BetaManagedAgentsAgentToolset20260401 object`55265527              - `type: "agent_toolset_20260401"`55285529              - `configs: array of BetaManagedAgentsAgentToolConfig`55305531                - `BetaManagedAgentsBashToolConfig object`55325533                  Configuration for the bash tool.55345535                  - `type: "bash"`55365537                  - `enabled: boolean`55385539                  - `name: "bash"`55405541                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`55425543                    Permission policy for tool execution.55445545                    - `BetaManagedAgentsAlwaysAllowPolicy object`55465547                      Tool calls are automatically approved without user confirmation.55485549                      - `type: "always_allow"`55505551                    - `BetaManagedAgentsAlwaysAskPolicy object`55525553                      Tool calls require user confirmation before execution.55545555                      - `type: "always_ask"`55565557                    - `BetaManagedAgentsAutoPolicy object`55585559                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.55605561                      - `type: "auto"`55625563                - `BetaManagedAgentsEditToolConfig object`55645565                  Configuration for the edit tool.55665567                  - `type: "edit"`55685569                  - `enabled: boolean`55705571                  - `name: "edit"`55725573                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`55745575                    Permission policy for tool execution.55765577                    - `BetaManagedAgentsAlwaysAllowPolicy object`55785579                      Tool calls are automatically approved without user confirmation.55805581                    - `BetaManagedAgentsAlwaysAskPolicy object`55825583                      Tool calls require user confirmation before execution.55845585                    - `BetaManagedAgentsAutoPolicy object`55865587                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.55885589                - `BetaManagedAgentsReadToolConfig object`55905591                  Configuration for the read tool.55925593                  - `type: "read"`55945595                  - `enabled: boolean`55965597                  - `name: "read"`55985599                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`56005601                    Permission policy for tool execution.56025603                    - `BetaManagedAgentsAlwaysAllowPolicy object`56045605                      Tool calls are automatically approved without user confirmation.56065607                    - `BetaManagedAgentsAlwaysAskPolicy object`56085609                      Tool calls require user confirmation before execution.56105611                    - `BetaManagedAgentsAutoPolicy object`56125613                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.56145615                - `BetaManagedAgentsWriteToolConfig object`56165617                  Configuration for the write tool.56185619                  - `type: "write"`56205621                  - `enabled: boolean`56225623                  - `name: "write"`56245625                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`56265627                    Permission policy for tool execution.56285629                    - `BetaManagedAgentsAlwaysAllowPolicy object`56305631                      Tool calls are automatically approved without user confirmation.56325633                    - `BetaManagedAgentsAlwaysAskPolicy object`56345635                      Tool calls require user confirmation before execution.56365637                    - `BetaManagedAgentsAutoPolicy object`56385639                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.56405641                - `BetaManagedAgentsGlobToolConfig object`56425643                  Configuration for the glob tool.56445645                  - `type: "glob"`56465647                  - `enabled: boolean`56485649                  - `name: "glob"`56505651                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`56525653                    Permission policy for tool execution.56545655                    - `BetaManagedAgentsAlwaysAllowPolicy object`56565657                      Tool calls are automatically approved without user confirmation.56585659                    - `BetaManagedAgentsAlwaysAskPolicy object`56605661                      Tool calls require user confirmation before execution.56625663                    - `BetaManagedAgentsAutoPolicy object`56645665                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.56665667                - `BetaManagedAgentsGrepToolConfig object`56685669                  Configuration for the grep tool.56705671                  - `type: "grep"`56725673                  - `enabled: boolean`56745675                  - `name: "grep"`56765677                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`56785679                    Permission policy for tool execution.56805681                    - `BetaManagedAgentsAlwaysAllowPolicy object`56825683                      Tool calls are automatically approved without user confirmation.56845685                    - `BetaManagedAgentsAlwaysAskPolicy object`56865687                      Tool calls require user confirmation before execution.56885689                    - `BetaManagedAgentsAutoPolicy object`56905691                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.56925693                - `BetaManagedAgentsWebFetchToolConfig object`56945695                  Configuration for the web_fetch tool.56965697                  - `type: "web_fetch"`56985699                  - `enabled: boolean`57005701                  - `name: "web_fetch"`57025703                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`57045705                    Permission policy for tool execution.57065707                    - `BetaManagedAgentsAlwaysAllowPolicy object`57085709                      Tool calls are automatically approved without user confirmation.57105711                    - `BetaManagedAgentsAlwaysAskPolicy object`57125713                      Tool calls require user confirmation before execution.57145715                    - `BetaManagedAgentsAutoPolicy object`57165717                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.57185719                  - `allowed_domains: optional array of string`57205721                  - `blocked_domains: optional array of string`57225723                  - `max_content_tokens: optional number or null`57245725                    format: int3257265727                - `BetaManagedAgentsWebSearchToolConfig object`57285729                  Configuration for the web_search tool.57305731                  - `type: "web_search"`57325733                  - `enabled: boolean`57345735                  - `name: "web_search"`57365737                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`57385739                    Permission policy for tool execution.57405741                    - `BetaManagedAgentsAlwaysAllowPolicy object`57425743                      Tool calls are automatically approved without user confirmation.57445745                    - `BetaManagedAgentsAlwaysAskPolicy object`57465747                      Tool calls require user confirmation before execution.57485749                    - `BetaManagedAgentsAutoPolicy object`57505751                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.57525753                  - `allowed_domains: optional array of string`57545755                  - `blocked_domains: optional array of string`57565757                  - `user_location: optional BetaManagedAgentsUserLocation or null`57585759                    Approximate user location for search result localization.57605761                    - `type: "approximate"`57625763                      Location precision. Only "approximate" is supported.57645765                    - `city: optional string or null`57665767                      City name.57685769                      minLength: 1, maxLength: 25557705771                    - `country: optional string or null`57725773                      Two-letter ISO 3166-1 country code, uppercase.57745775                    - `region: optional string or null`57765777                      Region or state name.57785779                      minLength: 1, maxLength: 25557805781                    - `timezone: optional string or null`57825783                      IANA timezone identifier, e.g. "America/Los_Angeles".57845785                      minLength: 1, maxLength: 25557865787              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`57885789                Resolved default configuration for agent tools.57905791                - `enabled: boolean`57925793                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`57945795                  Permission policy for tool execution.57965797                  - `BetaManagedAgentsAlwaysAllowPolicy object`57985799                    Tool calls are automatically approved without user confirmation.58005801                  - `BetaManagedAgentsAlwaysAskPolicy object`58025803                    Tool calls require user confirmation before execution.58045805                  - `BetaManagedAgentsAutoPolicy object`58065807                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.58085809            - `BetaManagedAgentsMCPToolset object`58105811              - `type: "mcp_toolset"`58125813              - `configs: array of BetaManagedAgentsMCPToolConfig`58145815                - `enabled: boolean`58165817                - `name: string`58185819                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`58205821                  Permission policy for tool execution.58225823                  - `BetaManagedAgentsAlwaysAllowPolicy object`58245825                    Tool calls are automatically approved without user confirmation.58265827                  - `BetaManagedAgentsAlwaysAskPolicy object`58285829                    Tool calls require user confirmation before execution.58305831                  - `BetaManagedAgentsAutoPolicy object`58325833                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.58345835              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`58365837                Resolved default configuration for all tools from an MCP server.58385839                - `enabled: boolean`58405841                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`58425843                  Permission policy for tool execution.58445845                  - `BetaManagedAgentsAlwaysAllowPolicy object`58465847                    Tool calls are automatically approved without user confirmation.58485849                  - `BetaManagedAgentsAlwaysAskPolicy object`58505851                    Tool calls require user confirmation before execution.58525853                  - `BetaManagedAgentsAutoPolicy object`58545855                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.58565857              - `mcp_server_name: string`58585859            - `BetaManagedAgentsCustomTool object`58605861              A custom tool as returned in API responses.58625863              - `type: "custom"`58645865              - `description: string`58665867              - `input_schema: BetaManagedAgentsCustomToolInputSchema`58685869                JSON Schema for custom tool input parameters.58705871                - `type: "object"`58725873                - `properties: optional map[unknown] or null`58745875                - `required: optional array of string or null`58765877              - `name: string`58785879          - `version: number`58805881            format: int3258825883        - `BetaManagedAgentsAdvisor object`58845885          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.58865887          - `type: "advisor"`58885889          - `model: string`58905891            The advisor model id.58925893    - `name: string`58945895    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`58965897      - `BetaManagedAgentsAnthropicSkill object`58985899        A resolved Anthropic-managed skill.59005901      - `BetaManagedAgentsCustomSkill object`59025903        A resolved user-created custom skill.59045905    - `system: string or null`59065907    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`59085909      - `BetaManagedAgentsAgentToolset20260401 object`59105911      - `BetaManagedAgentsMCPToolset object`59125913      - `BetaManagedAgentsCustomTool object`59145915        A custom tool as returned in API responses.59165917    - `version: number`59185919      format: int3259205921  - `archived_at: string or null`59225923    A timestamp in RFC 3339 format59245925    format: date-time59265927  - `budget: BetaManagedAgentsBudgetLimit or null`59285929    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.59305931    - `type: "limit"`59325933    - `max_list_cost: BetaMonetaryAmount`59345935      A monetary amount in a specific currency.59365937      - `amount: string`59385939        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.59405941      - `currency: BetaCurrency`59425943        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.59445945  - `created_at: string`59465947    A timestamp in RFC 3339 format59485949    format: date-time59505951  - `environment_id: string`59525953  - `metadata: map[string]`59545955  - `outcome_evaluations: array of BetaManagedAgentsOutcomeEvaluationResource`59565957    Per-outcome evaluation state. One entry per `define_outcome` event sent to the session.59585959    - `type: "outcome_evaluation"`59605961    - `completed_at: string or null`59625963      A timestamp in RFC 3339 format59645965      format: date-time59665967    - `description: string`59685969      What the agent should produce.59705971    - `explanation: string or null`59725973      Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.59745975    - `iteration: number`59765977      0-indexed revision cycle the outcome is currently on.59785979      format: int3259805981    - `outcome_id: string`59825983      Server-generated outc_ ID for this outcome.59845985    - `result: string`59865987      Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.59885989  - `resources: array of BetaManagedAgentsSessionResource`59905991    - `BetaManagedAgentsGitHubRepositoryResource object`59925993      - `type: "github_repository"`59945995      - `id: string`59965997      - `created_at: string`59985999        A timestamp in RFC 3339 format60006001        format: date-time60026003      - `mount_path: string`60046005      - `updated_at: string`60066007        A timestamp in RFC 3339 format60086009        format: date-time60106011      - `url: string`60126013      - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`60146015        - `BetaManagedAgentsBranchCheckout object`60166017          - `type: "branch"`60186019          - `name: string`60206021            Branch name to check out.60226023            minLength: 1, maxLength: 25560246025        - `BetaManagedAgentsCommitCheckout object`60266027          - `type: "commit"`60286029          - `sha: string`60306031            Full commit SHA to check out.60326033            minLength: 7, maxLength: 6460346035    - `BetaManagedAgentsFileResource object`60366037      - `type: "file"`60386039      - `id: string`60406041      - `created_at: string`60426043        A timestamp in RFC 3339 format60446045        format: date-time60466047      - `file_id: string`60486049      - `mount_path: string`60506051      - `updated_at: string`60526053        A timestamp in RFC 3339 format60546055        format: date-time60566057    - `BetaManagedAgentsMemoryStoreResource object`60586059      A memory store attached to an agent session.60606061      - `type: "memory_store"`60626063      - `memory_store_id: string`60646065        The memory store ID (memstore_...). Must belong to the caller's organization and workspace.60666067      - `access: optional "read_write" or "read_only" or null`60686069        Access mode for an attached memory store.60706071        - `"read_write"`60726073        - `"read_only"`60746075      - `description: optional string`60766077        Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.60786079      - `instructions: optional string or null`60806081        Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.60826083        maxLength: 409660846085      - `mount_path: optional string or null`60866087        Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.60886089      - `name: optional string or null`60906091        Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.60926093  - `stats: BetaManagedAgentsSessionStats`60946095    Timing statistics for a session.60966097    - `active_seconds: optional number`60986099      Cumulative time in seconds the session spent in `running` status. Excludes idle time.61006101      format: double61026103    - `duration_seconds: optional number`61046105      Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.61066107      format: double61086109  - `status: "rescheduling" or "running" or "idle" or "terminated"`61106111    SessionStatus enum61126113    - `"rescheduling"`61146115    - `"running"`61166117    - `"idle"`61186119    - `"terminated"`61206121  - `title: string or null`61226123  - `updated_at: string`61246125    A timestamp in RFC 3339 format61266127    format: date-time61286129  - `usage: BetaManagedAgentsSessionUsage`61306131    Cumulative token usage for a session across all turns.61326133    - `active_seconds: optional number`61346135      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.61366137      format: double61386139    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`61406141      Prompt-cache creation token usage broken down by cache lifetime.61426143      - `ephemeral_1h_input_tokens: optional number`61446145        Tokens used to create 1-hour ephemeral cache entries.61466147        format: int3261486149      - `ephemeral_5m_input_tokens: optional number`61506151        Tokens used to create 5-minute ephemeral cache entries.61526153        format: int3261546155    - `cache_read_input_tokens: optional number`61566157      Total tokens read from prompt cache.61586159      format: int3261606161    - `input_tokens: optional number`61626163      Total input tokens consumed across all turns.61646165      format: int3261666167    - `list_cost: optional BetaMonetaryAmount or null`61686169      A monetary amount in a specific currency.61706171    - `output_tokens: optional number`61726173      Total output tokens generated across all turns.61746175      format: int3261766177    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`61786179      Cumulative count of server-executed tool invocations, broken down by tool.61806181      - `web_fetch_requests: optional number`61826183        Number of server-executed web fetch requests.61846185        format: int3261866187      - `web_search_requests: optional number`61886189        Number of server-executed web search requests.61906191        format: int3261926193  - `vault_ids: array of string`61946195    Vault IDs attached to the session at creation. Empty when no vaults were supplied.61966197  - `deployment_id: optional string or null`61986199    Deployment ID when the session was created from a deployment reference. Null otherwise.62006201### Example62026203```bash6204curl https://api.anthropic.com/v1/sessions/$SESSION_ID \6205    -H 'Content-Type: application/json' \6206    -H 'anthropic-version: 2023-06-01' \6207    -H 'anthropic-beta: managed-agents-2026-04-01' \6208    -H "X-Api-Key: $ANTHROPIC_API_KEY" \6209    -d '{6210          "title": "Order #1234 inquiry"6211        }'6212```62136214#### Response (200)62156216```json6217{6218  "id": "sesn_011CZkZAtmR3yMPDzynEDxu7",6219  "agent": {6220    "id": "agent_011CZkYpogX7uDKUyvBTophP",6221    "description": "A general-purpose starter agent.",6222    "mcp_servers": [6223      {6224        "name": "example-mcp",6225        "type": "url",6226        "url": "https://example-server.modelcontextprotocol.io/sse"6227      }6228    ],6229    "model": {6230      "id": "claude-opus-5",6231      "effort": {6232        "type": "low"6233      },6234      "inference_geo": "inference_geo",6235      "speed": "standard"6236    },6237    "multiagent": {6238      "agents": [6239        {6240          "id": "agent_011CZkYqphY8vELVzwCUpqiQ",6241          "description": "A focused research subagent.",6242          "mcp_servers": [6243            {6244              "name": "example-mcp",6245              "type": "url",6246              "url": "https://example-server.modelcontextprotocol.io/sse"6247            }6248          ],6249          "model": {6250            "id": "claude-opus-5",6251            "effort": {6252              "type": "low"6253            },6254            "inference_geo": "inference_geo",6255            "speed": "standard"6256          },6257          "name": "Researcher",6258          "skills": [6259            {6260              "skill_id": "xlsx",6261              "type": "anthropic",6262              "version": "1"6263            }6264          ],6265          "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",6266          "tools": [6267            {6268              "configs": [6269                {6270                  "enabled": true,6271                  "name": "bash",6272                  "permission_policy": {6273                    "type": "always_allow"6274                  },6275                  "type": "bash"6276                }6277              ],6278              "default_config": {6279                "enabled": true,6280                "permission_policy": {6281                  "type": "always_ask"6282                }6283              },6284              "type": "agent_toolset_20260401"6285            }6286          ],6287          "type": "agent",6288          "version": 16289        }6290      ],6291      "type": "coordinator"6292    },6293    "name": "My First Agent",6294    "skills": [6295      {6296        "skill_id": "xlsx",6297        "type": "anthropic",6298        "version": "1"6299      },6300      {6301        "skill_id": "skill_011CZkZFNu9hAbo3jZPRgTlx",6302        "type": "custom",6303        "version": "2"6304      }6305    ],6306    "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user's task end to end.",6307    "tools": [6308      {6309        "configs": [6310          {6311            "enabled": true,6312            "name": "bash",6313            "permission_policy": {6314              "type": "always_allow"6315            },6316            "type": "bash"6317          }6318        ],6319        "default_config": {6320          "enabled": true,6321          "permission_policy": {6322            "type": "always_ask"6323          }6324        },6325        "type": "agent_toolset_20260401"6326      }6327    ],6328    "type": "agent",6329    "version": 16330  },6331  "archived_at": null,6332  "budget": {6333    "max_list_cost": {6334      "amount": "2500",6335      "currency": "USD"6336    },6337    "type": "limit"6338  },6339  "created_at": "2026-03-15T10:00:00Z",6340  "environment_id": "env_011CZkZ9X2dpNyB7HsEFoRfW",6341  "metadata": {},6342  "outcome_evaluations": [6343    {6344      "completed_at": "2026-03-15T10:02:31Z",6345      "description": "Produce a 2-page summary as summary.md",6346      "explanation": "All five sections present with inline citations.",6347      "iteration": 0,6348      "outcome_id": "outc_011CZkZRSw2kEfs6ncTVljxP",6349      "result": "satisfied",6350      "type": "outcome_evaluation"6351    }6352  ],6353  "resources": [6354    {6355      "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",6356      "created_at": "2026-03-15T10:00:00Z",6357      "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",6358      "mount_path": "/uploads/receipt.pdf",6359      "type": "file",6360      "updated_at": "2026-03-15T10:00:00Z"6361    },6362    {6363      "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",6364      "created_at": "2026-03-15T10:00:00Z",6365      "mount_path": "/workspace/example-repo",6366      "type": "github_repository",6367      "updated_at": "2026-03-15T10:00:00Z",6368      "url": "https://github.com/example-org/example-repo",6369      "checkout": {6370        "name": "main",6371        "type": "branch"6372      }6373    }6374  ],6375  "stats": {6376    "active_seconds": 0,6377    "duration_seconds": 06378  },6379  "status": "idle",6380  "title": "Order #1234 inquiry",6381  "type": "session",6382  "updated_at": "2026-03-15T10:00:00Z",6383  "usage": {6384    "active_seconds": 0,6385    "cache_creation": {6386      "ephemeral_1h_input_tokens": 0,6387      "ephemeral_5m_input_tokens": 06388    },6389    "cache_read_input_tokens": 0,6390    "input_tokens": 0,6391    "list_cost": {6392      "amount": "2500",6393      "currency": "USD"6394    },6395    "output_tokens": 0,6396    "server_tool_use": {6397      "web_fetch_requests": 0,6398      "web_search_requests": 36399    }6400  },6401  "vault_ids": [6402    "vlt_011CZkZDLs7fYzm1hXNPeRjv"6403  ],6404  "deployment_id": "deployment_id"6405}6406```64076408## Delete Session64096410**DELETE** `/v1/sessions/{session_id}`64116412Delete Session64136414### Path parameters64156416- `session_id: string`64176418### Headers64196420- `"anthropic-beta": optional array of AnthropicBeta`64216422  Optional header to specify the beta version(s) you want to use.64236424  - `string`64256426  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`64276428    - `"message-batches-2024-09-24"`64296430    - `"prompt-caching-2024-07-31"`64316432    - `"computer-use-2024-10-22"`64336434    - `"computer-use-2025-01-24"`64356436    - `"pdfs-2024-09-25"`64376438    - `"token-counting-2024-11-01"`64396440    - `"token-efficient-tools-2025-02-19"`64416442    - `"output-128k-2025-02-19"`64436444    - `"files-api-2025-04-14"`64456446    - `"mcp-client-2025-04-04"`64476448    - `"mcp-client-2025-11-20"`64496450    - `"dev-full-thinking-2025-05-14"`64516452    - `"interleaved-thinking-2025-05-14"`64536454    - `"code-execution-2025-05-22"`64556456    - `"extended-cache-ttl-2025-04-11"`64576458    - `"context-1m-2025-08-07"`64596460    - `"context-management-2025-06-27"`64616462    - `"model-context-window-exceeded-2025-08-26"`64636464    - `"skills-2025-10-02"`64656466    - `"fast-mode-2026-02-01"`64676468    - `"output-300k-2026-03-24"`64696470    - `"user-profiles-2026-03-24"`64716472    - `"user-profiles-2026-08-18"`64736474    - `"user-profiles-2026-09-04"`64756476    - `"advisor-tool-2026-03-01"`64776478    - `"managed-agents-2026-04-01"`64796480    - `"cache-diagnosis-2026-04-07"`64816482    - `"dreaming-2026-04-21"`64836484    - `"thinking-token-count-2026-05-13"`64856486    - `"server-side-fallback-2026-06-01"`64876488    - `"server-side-fallback-2026-07-01"`64896490    - `"fallback-credit-2026-06-01"`64916492    - `"fallback-credit-2026-07-01"`64936494    - `"agent-memory-2026-07-22"`64956496    - `"mid-conversation-tool-changes-2026-07-01"`64976498    - `"compact-2026-01-12"`64996500    - `"computer-use-2025-11-24"`65016502    - `"mcp-tunnels-2026-06-22"`65036504    - `"structured-outputs-2025-11-13"`65056506    - `"task-budgets-2026-03-13"`65076508    - `"thinking-display-updates-2026-08-18"`65096510    - `"ce-user-management-2026-07-13"`65116512    - `"mid-conversation-output-config-2026-07-01"`65136514    - `"thinking-binding-controls-2026-08-01"`65156516    - `"mid-conversation-system-clear-at-2026-08-21"`65176518    - `"compact-2026-09-04"`65196520- `"anthropic-workspace-id": optional string`65216522### Returns65236524- `BetaManagedAgentsDeletedSession object`65256526  Confirmation that a `session` has been permanently deleted.65276528  - `type: "session_deleted"`65296530  - `id: string`65316532### Example65336534```bash6535curl https://api.anthropic.com/v1/sessions/$SESSION_ID \6536    -X DELETE \6537    -H 'anthropic-version: 2023-06-01' \6538    -H 'anthropic-beta: managed-agents-2026-04-01' \6539    -H "X-Api-Key: $ANTHROPIC_API_KEY"6540```65416542#### Response (200)65436544```json6545{6546  "id": "sesn_011CZkZAtmR3yMPDzynEDxu7",6547  "type": "session_deleted"6548}6549```65506551## Archive Session65526553**POST** `/v1/sessions/{session_id}/archive`65546555Archive Session65566557### Path parameters65586559- `session_id: string`65606561### Headers65626563- `"anthropic-beta": optional array of AnthropicBeta`65646565  Optional header to specify the beta version(s) you want to use.65666567  - `string`65686569  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`65706571    - `"message-batches-2024-09-24"`65726573    - `"prompt-caching-2024-07-31"`65746575    - `"computer-use-2024-10-22"`65766577    - `"computer-use-2025-01-24"`65786579    - `"pdfs-2024-09-25"`65806581    - `"token-counting-2024-11-01"`65826583    - `"token-efficient-tools-2025-02-19"`65846585    - `"output-128k-2025-02-19"`65866587    - `"files-api-2025-04-14"`65886589    - `"mcp-client-2025-04-04"`65906591    - `"mcp-client-2025-11-20"`65926593    - `"dev-full-thinking-2025-05-14"`65946595    - `"interleaved-thinking-2025-05-14"`65966597    - `"code-execution-2025-05-22"`65986599    - `"extended-cache-ttl-2025-04-11"`66006601    - `"context-1m-2025-08-07"`66026603    - `"context-management-2025-06-27"`66046605    - `"model-context-window-exceeded-2025-08-26"`66066607    - `"skills-2025-10-02"`66086609    - `"fast-mode-2026-02-01"`66106611    - `"output-300k-2026-03-24"`66126613    - `"user-profiles-2026-03-24"`66146615    - `"user-profiles-2026-08-18"`66166617    - `"user-profiles-2026-09-04"`66186619    - `"advisor-tool-2026-03-01"`66206621    - `"managed-agents-2026-04-01"`66226623    - `"cache-diagnosis-2026-04-07"`66246625    - `"dreaming-2026-04-21"`66266627    - `"thinking-token-count-2026-05-13"`66286629    - `"server-side-fallback-2026-06-01"`66306631    - `"server-side-fallback-2026-07-01"`66326633    - `"fallback-credit-2026-06-01"`66346635    - `"fallback-credit-2026-07-01"`66366637    - `"agent-memory-2026-07-22"`66386639    - `"mid-conversation-tool-changes-2026-07-01"`66406641    - `"compact-2026-01-12"`66426643    - `"computer-use-2025-11-24"`66446645    - `"mcp-tunnels-2026-06-22"`66466647    - `"structured-outputs-2025-11-13"`66486649    - `"task-budgets-2026-03-13"`66506651    - `"thinking-display-updates-2026-08-18"`66526653    - `"ce-user-management-2026-07-13"`66546655    - `"mid-conversation-output-config-2026-07-01"`66566657    - `"thinking-binding-controls-2026-08-01"`66586659    - `"mid-conversation-system-clear-at-2026-08-21"`66606661    - `"compact-2026-09-04"`66626663- `"anthropic-workspace-id": optional string`66646665### Returns66666667- `BetaManagedAgentsSession object`66686669  A Managed Agents `session`.66706671  - `type: "session"`66726673  - `id: string`66746675  - `agent: BetaManagedAgentsSessionAgent`66766677    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.66786679    - `type: "agent"`66806681    - `id: string`66826683    - `description: string or null`66846685    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`66866687      - `type: "url"`66886689      - `name: string`66906691      - `url: string`66926693    - `model: BetaManagedAgentsModelConfig`66946695      Model identifier and configuration.66966697      - `id: BetaManagedAgentsModel`66986699        The model that will power your agent.67006701        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.67026703        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`67046705          The model that will power your agent.67066707          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.67086709          - `"claude-fable-5-1"`67106711            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows67126713          - `"claude-sonnet-5"`67146715            High-performance model for coding and agents67166717          - `"claude-fable-5"`67186719            Next generation of intelligence for the hardest knowledge work and coding problems67206721          - `"claude-opus-5"`67226723            Powerful intelligence for long-running agents and coding67246725          - `"claude-opus-4-8"`67266727            Powerful intelligence for long-running agents and coding67286729          - `"claude-opus-4-7"`67306731            Powerful intelligence for long-running agents and coding67326733          - `"claude-opus-4-6"`67346735            Powerful intelligence for long-running agents and coding67366737          - `"claude-sonnet-4-6"`67386739            Best combination of speed and intelligence67406741          - `"claude-haiku-4-5"`67426743            Fastest model with near-frontier intelligence67446745          - `"claude-haiku-4-5-20251001"`67466747            Fastest model with near-frontier intelligence67486749          - `"claude-opus-4-5"`67506751            Powerful intelligence for long-running agents and coding67526753          - `"claude-opus-4-5-20251101"`67546755            Powerful intelligence for long-running agents and coding67566757          - `"claude-sonnet-4-5"`67586759            High-performance model for agents and coding67606761          - `"claude-sonnet-4-5-20250929"`67626763            High-performance model for agents and coding67646765        - `string`67666767      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`67686769        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.67706771        - `BetaManagedAgentsEffortLow object`67726773          Low effort. Favors latency over reasoning depth.67746775          - `type: "low"`67766777        - `BetaManagedAgentsEffortMedium object`67786779          Medium effort. Balances latency and reasoning depth.67806781          - `type: "medium"`67826783        - `BetaManagedAgentsEffortHigh object`67846785          High effort. Favors reasoning depth.67866787          - `type: "high"`67886789        - `BetaManagedAgentsEffortXhigh object`67906791          Extra-high effort. Not all models accept this level.67926793          - `type: "xhigh"`67946795        - `BetaManagedAgentsEffortMax object`67966797          Maximum effort. Favors reasoning depth over latency.67986799          - `type: "max"`68006801      - `inference_geo: optional string`68026803        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.68046805      - `speed: optional "standard" or "fast"`68066807        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.68086809        - `"standard"`68106811        - `"fast"`68126813    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`68146815      Resolved coordinator topology with full agent definitions for each roster member.68166817      - `type: "coordinator"`68186819      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`68206821        Full `agent` definitions the coordinator may spawn as session threads.68226823        - `BetaManagedAgentsSessionThreadAgent object`68246825          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.68266827          - `type: "agent"`68286829          - `id: string`68306831          - `description: string or null`68326833          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`68346835            - `type: "url"`68366837            - `name: string`68386839            - `url: string`68406841          - `model: BetaManagedAgentsModelConfig`68426843            Model identifier and configuration.68446845          - `name: string`68466847          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`68486849            - `BetaManagedAgentsAnthropicSkill object`68506851              A resolved Anthropic-managed skill.68526853              - `type: "anthropic"`68546855              - `skill_id: string`68566857              - `version: string`68586859            - `BetaManagedAgentsCustomSkill object`68606861              A resolved user-created custom skill.68626863              - `type: "custom"`68646865              - `skill_id: string`68666867              - `version: string`68686869          - `system: string or null`68706871          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`68726873            - `BetaManagedAgentsAgentToolset20260401 object`68746875              - `type: "agent_toolset_20260401"`68766877              - `configs: array of BetaManagedAgentsAgentToolConfig`68786879                - `BetaManagedAgentsBashToolConfig object`68806881                  Configuration for the bash tool.68826883                  - `type: "bash"`68846885                  - `enabled: boolean`68866887                  - `name: "bash"`68886889                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`68906891                    Permission policy for tool execution.68926893                    - `BetaManagedAgentsAlwaysAllowPolicy object`68946895                      Tool calls are automatically approved without user confirmation.68966897                      - `type: "always_allow"`68986899                    - `BetaManagedAgentsAlwaysAskPolicy object`69006901                      Tool calls require user confirmation before execution.69026903                      - `type: "always_ask"`69046905                    - `BetaManagedAgentsAutoPolicy object`69066907                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.69086909                      - `type: "auto"`69106911                - `BetaManagedAgentsEditToolConfig object`69126913                  Configuration for the edit tool.69146915                  - `type: "edit"`69166917                  - `enabled: boolean`69186919                  - `name: "edit"`69206921                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`69226923                    Permission policy for tool execution.69246925                    - `BetaManagedAgentsAlwaysAllowPolicy object`69266927                      Tool calls are automatically approved without user confirmation.69286929                    - `BetaManagedAgentsAlwaysAskPolicy object`69306931                      Tool calls require user confirmation before execution.69326933                    - `BetaManagedAgentsAutoPolicy object`69346935                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.69366937                - `BetaManagedAgentsReadToolConfig object`69386939                  Configuration for the read tool.69406941                  - `type: "read"`69426943                  - `enabled: boolean`69446945                  - `name: "read"`69466947                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`69486949                    Permission policy for tool execution.69506951                    - `BetaManagedAgentsAlwaysAllowPolicy object`69526953                      Tool calls are automatically approved without user confirmation.69546955                    - `BetaManagedAgentsAlwaysAskPolicy object`69566957                      Tool calls require user confirmation before execution.69586959                    - `BetaManagedAgentsAutoPolicy object`69606961                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.69626963                - `BetaManagedAgentsWriteToolConfig object`69646965                  Configuration for the write tool.69666967                  - `type: "write"`69686969                  - `enabled: boolean`69706971                  - `name: "write"`69726973                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`69746975                    Permission policy for tool execution.69766977                    - `BetaManagedAgentsAlwaysAllowPolicy object`69786979                      Tool calls are automatically approved without user confirmation.69806981                    - `BetaManagedAgentsAlwaysAskPolicy object`69826983                      Tool calls require user confirmation before execution.69846985                    - `BetaManagedAgentsAutoPolicy object`69866987                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.69886989                - `BetaManagedAgentsGlobToolConfig object`69906991                  Configuration for the glob tool.69926993                  - `type: "glob"`69946995                  - `enabled: boolean`69966997                  - `name: "glob"`69986999                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`70007001                    Permission policy for tool execution.70027003                    - `BetaManagedAgentsAlwaysAllowPolicy object`70047005                      Tool calls are automatically approved without user confirmation.70067007                    - `BetaManagedAgentsAlwaysAskPolicy object`70087009                      Tool calls require user confirmation before execution.70107011                    - `BetaManagedAgentsAutoPolicy object`70127013                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.70147015                - `BetaManagedAgentsGrepToolConfig object`70167017                  Configuration for the grep tool.70187019                  - `type: "grep"`70207021                  - `enabled: boolean`70227023                  - `name: "grep"`70247025                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`70267027                    Permission policy for tool execution.70287029                    - `BetaManagedAgentsAlwaysAllowPolicy object`70307031                      Tool calls are automatically approved without user confirmation.70327033                    - `BetaManagedAgentsAlwaysAskPolicy object`70347035                      Tool calls require user confirmation before execution.70367037                    - `BetaManagedAgentsAutoPolicy object`70387039                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.70407041                - `BetaManagedAgentsWebFetchToolConfig object`70427043                  Configuration for the web_fetch tool.70447045                  - `type: "web_fetch"`70467047                  - `enabled: boolean`70487049                  - `name: "web_fetch"`70507051                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`70527053                    Permission policy for tool execution.70547055                    - `BetaManagedAgentsAlwaysAllowPolicy object`70567057                      Tool calls are automatically approved without user confirmation.70587059                    - `BetaManagedAgentsAlwaysAskPolicy object`70607061                      Tool calls require user confirmation before execution.70627063                    - `BetaManagedAgentsAutoPolicy object`70647065                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.70667067                  - `allowed_domains: optional array of string`70687069                  - `blocked_domains: optional array of string`70707071                  - `max_content_tokens: optional number or null`70727073                    format: int3270747075                - `BetaManagedAgentsWebSearchToolConfig object`70767077                  Configuration for the web_search tool.70787079                  - `type: "web_search"`70807081                  - `enabled: boolean`70827083                  - `name: "web_search"`70847085                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`70867087                    Permission policy for tool execution.70887089                    - `BetaManagedAgentsAlwaysAllowPolicy object`70907091                      Tool calls are automatically approved without user confirmation.70927093                    - `BetaManagedAgentsAlwaysAskPolicy object`70947095                      Tool calls require user confirmation before execution.70967097                    - `BetaManagedAgentsAutoPolicy object`70987099                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.71007101                  - `allowed_domains: optional array of string`71027103                  - `blocked_domains: optional array of string`71047105                  - `user_location: optional BetaManagedAgentsUserLocation or null`71067107                    Approximate user location for search result localization.71087109                    - `type: "approximate"`71107111                      Location precision. Only "approximate" is supported.71127113                    - `city: optional string or null`71147115                      City name.71167117                      minLength: 1, maxLength: 25571187119                    - `country: optional string or null`71207121                      Two-letter ISO 3166-1 country code, uppercase.71227123                    - `region: optional string or null`71247125                      Region or state name.71267127                      minLength: 1, maxLength: 25571287129                    - `timezone: optional string or null`71307131                      IANA timezone identifier, e.g. "America/Los_Angeles".71327133                      minLength: 1, maxLength: 25571347135              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`71367137                Resolved default configuration for agent tools.71387139                - `enabled: boolean`71407141                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`71427143                  Permission policy for tool execution.71447145                  - `BetaManagedAgentsAlwaysAllowPolicy object`71467147                    Tool calls are automatically approved without user confirmation.71487149                  - `BetaManagedAgentsAlwaysAskPolicy object`71507151                    Tool calls require user confirmation before execution.71527153                  - `BetaManagedAgentsAutoPolicy object`71547155                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.71567157            - `BetaManagedAgentsMCPToolset object`71587159              - `type: "mcp_toolset"`71607161              - `configs: array of BetaManagedAgentsMCPToolConfig`71627163                - `enabled: boolean`71647165                - `name: string`71667167                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`71687169                  Permission policy for tool execution.71707171                  - `BetaManagedAgentsAlwaysAllowPolicy object`71727173                    Tool calls are automatically approved without user confirmation.71747175                  - `BetaManagedAgentsAlwaysAskPolicy object`71767177                    Tool calls require user confirmation before execution.71787179                  - `BetaManagedAgentsAutoPolicy object`71807181                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.71827183              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`71847185                Resolved default configuration for all tools from an MCP server.71867187                - `enabled: boolean`71887189                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`71907191                  Permission policy for tool execution.71927193                  - `BetaManagedAgentsAlwaysAllowPolicy object`71947195                    Tool calls are automatically approved without user confirmation.71967197                  - `BetaManagedAgentsAlwaysAskPolicy object`71987199                    Tool calls require user confirmation before execution.72007201                  - `BetaManagedAgentsAutoPolicy object`72027203                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.72047205              - `mcp_server_name: string`72067207            - `BetaManagedAgentsCustomTool object`72087209              A custom tool as returned in API responses.72107211              - `type: "custom"`72127213              - `description: string`72147215              - `input_schema: BetaManagedAgentsCustomToolInputSchema`72167217                JSON Schema for custom tool input parameters.72187219                - `type: "object"`72207221                - `properties: optional map[unknown] or null`72227223                - `required: optional array of string or null`72247225              - `name: string`72267227          - `version: number`72287229            format: int3272307231        - `BetaManagedAgentsAdvisor object`72327233          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.72347235          - `type: "advisor"`72367237          - `model: string`72387239            The advisor model id.72407241    - `name: string`72427243    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`72447245      - `BetaManagedAgentsAnthropicSkill object`72467247        A resolved Anthropic-managed skill.72487249      - `BetaManagedAgentsCustomSkill object`72507251        A resolved user-created custom skill.72527253    - `system: string or null`72547255    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`72567257      - `BetaManagedAgentsAgentToolset20260401 object`72587259      - `BetaManagedAgentsMCPToolset object`72607261      - `BetaManagedAgentsCustomTool object`72627263        A custom tool as returned in API responses.72647265    - `version: number`72667267      format: int3272687269  - `archived_at: string or null`72707271    A timestamp in RFC 3339 format72727273    format: date-time72747275  - `budget: BetaManagedAgentsBudgetLimit or null`72767277    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.72787279    - `type: "limit"`72807281    - `max_list_cost: BetaMonetaryAmount`72827283      A monetary amount in a specific currency.72847285      - `amount: string`72867287        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.72887289      - `currency: BetaCurrency`72907291        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.72927293  - `created_at: string`72947295    A timestamp in RFC 3339 format72967297    format: date-time72987299  - `environment_id: string`73007301  - `metadata: map[string]`73027303  - `outcome_evaluations: array of BetaManagedAgentsOutcomeEvaluationResource`73047305    Per-outcome evaluation state. One entry per `define_outcome` event sent to the session.73067307    - `type: "outcome_evaluation"`73087309    - `completed_at: string or null`73107311      A timestamp in RFC 3339 format73127313      format: date-time73147315    - `description: string`73167317      What the agent should produce.73187319    - `explanation: string or null`73207321      Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.73227323    - `iteration: number`73247325      0-indexed revision cycle the outcome is currently on.73267327      format: int3273287329    - `outcome_id: string`73307331      Server-generated outc_ ID for this outcome.73327333    - `result: string`73347335      Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.73367337  - `resources: array of BetaManagedAgentsSessionResource`73387339    - `BetaManagedAgentsGitHubRepositoryResource object`73407341      - `type: "github_repository"`73427343      - `id: string`73447345      - `created_at: string`73467347        A timestamp in RFC 3339 format73487349        format: date-time73507351      - `mount_path: string`73527353      - `updated_at: string`73547355        A timestamp in RFC 3339 format73567357        format: date-time73587359      - `url: string`73607361      - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`73627363        - `BetaManagedAgentsBranchCheckout object`73647365          - `type: "branch"`73667367          - `name: string`73687369            Branch name to check out.73707371            minLength: 1, maxLength: 25573727373        - `BetaManagedAgentsCommitCheckout object`73747375          - `type: "commit"`73767377          - `sha: string`73787379            Full commit SHA to check out.73807381            minLength: 7, maxLength: 6473827383    - `BetaManagedAgentsFileResource object`73847385      - `type: "file"`73867387      - `id: string`73887389      - `created_at: string`73907391        A timestamp in RFC 3339 format73927393        format: date-time73947395      - `file_id: string`73967397      - `mount_path: string`73987399      - `updated_at: string`74007401        A timestamp in RFC 3339 format74027403        format: date-time74047405    - `BetaManagedAgentsMemoryStoreResource object`74067407      A memory store attached to an agent session.74087409      - `type: "memory_store"`74107411      - `memory_store_id: string`74127413        The memory store ID (memstore_...). Must belong to the caller's organization and workspace.74147415      - `access: optional "read_write" or "read_only" or null`74167417        Access mode for an attached memory store.74187419        - `"read_write"`74207421        - `"read_only"`74227423      - `description: optional string`74247425        Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.74267427      - `instructions: optional string or null`74287429        Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.74307431        maxLength: 409674327433      - `mount_path: optional string or null`74347435        Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.74367437      - `name: optional string or null`74387439        Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.74407441  - `stats: BetaManagedAgentsSessionStats`74427443    Timing statistics for a session.74447445    - `active_seconds: optional number`74467447      Cumulative time in seconds the session spent in `running` status. Excludes idle time.74487449      format: double74507451    - `duration_seconds: optional number`74527453      Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.74547455      format: double74567457  - `status: "rescheduling" or "running" or "idle" or "terminated"`74587459    SessionStatus enum74607461    - `"rescheduling"`74627463    - `"running"`74647465    - `"idle"`74667467    - `"terminated"`74687469  - `title: string or null`74707471  - `updated_at: string`74727473    A timestamp in RFC 3339 format74747475    format: date-time74767477  - `usage: BetaManagedAgentsSessionUsage`74787479    Cumulative token usage for a session across all turns.74807481    - `active_seconds: optional number`74827483      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.74847485      format: double74867487    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`74887489      Prompt-cache creation token usage broken down by cache lifetime.74907491      - `ephemeral_1h_input_tokens: optional number`74927493        Tokens used to create 1-hour ephemeral cache entries.74947495        format: int3274967497      - `ephemeral_5m_input_tokens: optional number`74987499        Tokens used to create 5-minute ephemeral cache entries.75007501        format: int3275027503    - `cache_read_input_tokens: optional number`75047505      Total tokens read from prompt cache.75067507      format: int3275087509    - `input_tokens: optional number`75107511      Total input tokens consumed across all turns.75127513      format: int3275147515    - `list_cost: optional BetaMonetaryAmount or null`75167517      A monetary amount in a specific currency.75187519    - `output_tokens: optional number`75207521      Total output tokens generated across all turns.75227523      format: int3275247525    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`75267527      Cumulative count of server-executed tool invocations, broken down by tool.75287529      - `web_fetch_requests: optional number`75307531        Number of server-executed web fetch requests.75327533        format: int3275347535      - `web_search_requests: optional number`75367537        Number of server-executed web search requests.75387539        format: int3275407541  - `vault_ids: array of string`75427543    Vault IDs attached to the session at creation. Empty when no vaults were supplied.75447545  - `deployment_id: optional string or null`75467547    Deployment ID when the session was created from a deployment reference. Null otherwise.75487549### Example75507551```bash7552curl https://api.anthropic.com/v1/sessions/$SESSION_ID/archive \7553    -X POST \7554    -H 'anthropic-version: 2023-06-01' \7555    -H 'anthropic-beta: managed-agents-2026-04-01' \7556    -H "X-Api-Key: $ANTHROPIC_API_KEY"7557```75587559#### Response (200)75607561```json7562{7563  "id": "sesn_011CZkZAtmR3yMPDzynEDxu7",7564  "agent": {7565    "id": "agent_011CZkYpogX7uDKUyvBTophP",7566    "description": "A general-purpose starter agent.",7567    "mcp_servers": [7568      {7569        "name": "example-mcp",7570        "type": "url",7571        "url": "https://example-server.modelcontextprotocol.io/sse"7572      }7573    ],7574    "model": {7575      "id": "claude-opus-5",7576      "effort": {7577        "type": "low"7578      },7579      "inference_geo": "inference_geo",7580      "speed": "standard"7581    },7582    "multiagent": {7583      "agents": [7584        {7585          "id": "agent_011CZkYqphY8vELVzwCUpqiQ",7586          "description": "A focused research subagent.",7587          "mcp_servers": [7588            {7589              "name": "example-mcp",7590              "type": "url",7591              "url": "https://example-server.modelcontextprotocol.io/sse"7592            }7593          ],7594          "model": {7595            "id": "claude-opus-5",7596            "effort": {7597              "type": "low"7598            },7599            "inference_geo": "inference_geo",7600            "speed": "standard"7601          },7602          "name": "Researcher",7603          "skills": [7604            {7605              "skill_id": "xlsx",7606              "type": "anthropic",7607              "version": "1"7608            }7609          ],7610          "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",7611          "tools": [7612            {7613              "configs": [7614                {7615                  "enabled": true,7616                  "name": "bash",7617                  "permission_policy": {7618                    "type": "always_allow"7619                  },7620                  "type": "bash"7621                }7622              ],7623              "default_config": {7624                "enabled": true,7625                "permission_policy": {7626                  "type": "always_ask"7627                }7628              },7629              "type": "agent_toolset_20260401"7630            }7631          ],7632          "type": "agent",7633          "version": 17634        }7635      ],7636      "type": "coordinator"7637    },7638    "name": "My First Agent",7639    "skills": [7640      {7641        "skill_id": "xlsx",7642        "type": "anthropic",7643        "version": "1"7644      },7645      {7646        "skill_id": "skill_011CZkZFNu9hAbo3jZPRgTlx",7647        "type": "custom",7648        "version": "2"7649      }7650    ],7651    "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user's task end to end.",7652    "tools": [7653      {7654        "configs": [7655          {7656            "enabled": true,7657            "name": "bash",7658            "permission_policy": {7659              "type": "always_allow"7660            },7661            "type": "bash"7662          }7663        ],7664        "default_config": {7665          "enabled": true,7666          "permission_policy": {7667            "type": "always_ask"7668          }7669        },7670        "type": "agent_toolset_20260401"7671      }7672    ],7673    "type": "agent",7674    "version": 17675  },7676  "archived_at": null,7677  "budget": {7678    "max_list_cost": {7679      "amount": "2500",7680      "currency": "USD"7681    },7682    "type": "limit"7683  },7684  "created_at": "2026-03-15T10:00:00Z",7685  "environment_id": "env_011CZkZ9X2dpNyB7HsEFoRfW",7686  "metadata": {},7687  "outcome_evaluations": [7688    {7689      "completed_at": "2026-03-15T10:02:31Z",7690      "description": "Produce a 2-page summary as summary.md",7691      "explanation": "All five sections present with inline citations.",7692      "iteration": 0,7693      "outcome_id": "outc_011CZkZRSw2kEfs6ncTVljxP",7694      "result": "satisfied",7695      "type": "outcome_evaluation"7696    }7697  ],7698  "resources": [7699    {7700      "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",7701      "created_at": "2026-03-15T10:00:00Z",7702      "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",7703      "mount_path": "/uploads/receipt.pdf",7704      "type": "file",7705      "updated_at": "2026-03-15T10:00:00Z"7706    },7707    {7708      "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",7709      "created_at": "2026-03-15T10:00:00Z",7710      "mount_path": "/workspace/example-repo",7711      "type": "github_repository",7712      "updated_at": "2026-03-15T10:00:00Z",7713      "url": "https://github.com/example-org/example-repo",7714      "checkout": {7715        "name": "main",7716        "type": "branch"7717      }7718    }7719  ],7720  "stats": {7721    "active_seconds": 0,7722    "duration_seconds": 07723  },7724  "status": "idle",7725  "title": "Order #1234 inquiry",7726  "type": "session",7727  "updated_at": "2026-03-15T10:00:00Z",7728  "usage": {7729    "active_seconds": 0,7730    "cache_creation": {7731      "ephemeral_1h_input_tokens": 0,7732      "ephemeral_5m_input_tokens": 07733    },7734    "cache_read_input_tokens": 0,7735    "input_tokens": 0,7736    "list_cost": {7737      "amount": "2500",7738      "currency": "USD"7739    },7740    "output_tokens": 0,7741    "server_tool_use": {7742      "web_fetch_requests": 0,7743      "web_search_requests": 37744    }7745  },7746  "vault_ids": [7747    "vlt_011CZkZDLs7fYzm1hXNPeRjv"7748  ],7749  "deployment_id": "deployment_id"7750}7751```77527753## Domain types77547755### Beta Managed Agents Advisor Params77567757- `BetaManagedAgentsAdvisorParams object`77587759  Platform advisor roster entry: a model the session's primary thread may consult mid-turn. At most one per roster; the entry occupies the roster name `anthropic.advisor`.77607761  - `type: "advisor"`77627763  - `model: string`77647765    A Claude model id. The model must be permitted as an advisor for this agent's model — see the sessions/threads/advisor spec.77667767    minLength: 1, maxLength: 25677687769### Beta Managed Agents Agent Message Preview77707771- `BetaManagedAgentsAgentMessagePreview object`77727773  - `type: "agent.message"`77747775  - `id: string`77767777    The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.77787779### Beta Managed Agents Agent Params77807781- `BetaManagedAgentsAgentParams object`77827783  Specification for an Agent. Provide a specific `version` or use the short-form `agent="agent_id"` for the most recent version77847785  - `type: "agent"`77867787  - `id: string`77887789    The `agent` ID.77907791    minLength: 1, maxLength: 12877927793  - `version: optional number`77947795    The specific `agent` version to use. Omit to use the latest version. Must be at least 1 if specified.77967797    format: int3277987799### Beta Managed Agents Agent Thinking Preview78007801- `BetaManagedAgentsAgentThinkingPreview object`78027803  - `type: "agent.thinking"`78047805  - `id: string`78067807    The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.78087809### Beta Managed Agents Agent With Overrides Params78107811- `BetaManagedAgentsAgentWithOverridesParams object`78127813  Reference to an `agent` plus optional configuration overrides. Each provided field replaces the agent's value for the caller's use; the agent resource is unchanged.78147815  - `type: "agent_with_overrides"`78167817  - `id: string`78187819    The `agent` ID.78207821    minLength: 1, maxLength: 12878227823  - `mcp_servers: optional array of BetaManagedAgentsURLMCPServerParams`78247825    Replacement MCP server list. Full replacement: the provided array becomes the MCP servers. Send an empty array to clear; omit to preserve the agent's servers.78267827    - `type: "url"`78287829    - `name: string`78307831      Unique name for this server, referenced by mcp_toolset configurations. 1-255 characters.78327833      minLength: 1, maxLength: 25578347835    - `url: string`78367837      Endpoint URL for the MCP server.78387839      maxLength: 204878407841  - `model: optional BetaManagedAgentsModel or BetaManagedAgentsModelConfigParams`78427843    Replacement model. Accepts the model string, e.g. `claude-opus-5`, or a `model_config` object. Omit to use the agent's model.78447845    - `BetaManagedAgentsModel = "claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more or string`78467847      The model that will power your agent.78487849      See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.78507851      - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`78527853        The model that will power your agent.78547855        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.78567857        - `"claude-fable-5-1"`78587859          Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows78607861        - `"claude-sonnet-5"`78627863          High-performance model for coding and agents78647865        - `"claude-fable-5"`78667867          Next generation of intelligence for the hardest knowledge work and coding problems78687869        - `"claude-opus-5"`78707871          Powerful intelligence for long-running agents and coding78727873        - `"claude-opus-4-8"`78747875          Powerful intelligence for long-running agents and coding78767877        - `"claude-opus-4-7"`78787879          Powerful intelligence for long-running agents and coding78807881        - `"claude-opus-4-6"`78827883          Powerful intelligence for long-running agents and coding78847885        - `"claude-sonnet-4-6"`78867887          Best combination of speed and intelligence78887889        - `"claude-haiku-4-5"`78907891          Fastest model with near-frontier intelligence78927893        - `"claude-haiku-4-5-20251001"`78947895          Fastest model with near-frontier intelligence78967897        - `"claude-opus-4-5"`78987899          Powerful intelligence for long-running agents and coding79007901        - `"claude-opus-4-5-20251101"`79027903          Powerful intelligence for long-running agents and coding79047905        - `"claude-sonnet-4-5"`79067907          High-performance model for agents and coding79087909        - `"claude-sonnet-4-5-20250929"`79107911          High-performance model for agents and coding79127913      - `string`79147915    - `BetaManagedAgentsModelConfigParams object`79167917      An object that defines additional configuration control over model use79187919      - `id: BetaManagedAgentsModel`79207921        The model that will power your agent.79227923        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.79247925      - `effort: optional "low" or "medium" or "high" or 2 more or BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or 3 more or null`79267927        How hard Claude works on each inference call. Accepts a bare level string (`"high"`) or `{"type": "high"}`. On create, omitting it resolves the per-model default; on update, omitting it leaves the stored value unchanged.79287929        - `BetaManagedAgentsEffortLevel = "low" or "medium" or "high" or 2 more`79307931          How hard Claude works on each turn. Higher levels favor reasoning depth over latency. Not all models accept every level; invalid combinations are rejected at create time.79327933          - `"low"`79347935          - `"medium"`79367937          - `"high"`79387939          - `"xhigh"`79407941          - `"max"`79427943        - `BetaManagedAgentsEffortLow object`79447945          Low effort. Favors latency over reasoning depth.79467947          - `type: "low"`79487949        - `BetaManagedAgentsEffortMedium object`79507951          Medium effort. Balances latency and reasoning depth.79527953          - `type: "medium"`79547955        - `BetaManagedAgentsEffortHigh object`79567957          High effort. Favors reasoning depth.79587959          - `type: "high"`79607961        - `BetaManagedAgentsEffortXhigh object`79627963          Extra-high effort. Not all models accept this level.79647965          - `type: "xhigh"`79667967        - `BetaManagedAgentsEffortMax object`79687969          Maximum effort. Favors reasoning depth over latency.79707971          - `type: "max"`79727973      - `inference_geo: optional string or null`79747975        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo. On update, `model` is whole-object replacement — omitting inference_geo clears it.79767977      - `speed: optional "standard" or "fast" or null`79787979        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.79807981        - `"standard"`79827983        - `"fast"`79847985  - `skills: optional array of BetaManagedAgentsSkillParams`79867987    Replacement skill list. Full replacement: the provided array becomes the skills. Send an empty array to clear; omit to preserve the agent's skills.79887989    - `BetaManagedAgentsAnthropicSkillParams object`79907991      An Anthropic-managed skill.79927993      - `type: "anthropic"`79947995      - `skill_id: string`79967997        Identifier of the Anthropic skill (e.g., "xlsx").79987999        minLength: 1, maxLength: 6480008001      - `version: optional string or null`80028003        Version to pin. Defaults to latest if omitted.80048005        minLength: 1, maxLength: 6480068007    - `BetaManagedAgentsCustomSkillParams object`80088009      A user-created custom skill.80108011      - `type: "custom"`80128013      - `skill_id: string`80148015        Tagged ID of the custom skill (e.g., "skill_01XJ5...").80168017        minLength: 1, maxLength: 6480188019      - `version: optional string or null`80208021        Version to pin. Defaults to latest if omitted.80228023        minLength: 1, maxLength: 6480248025  - `system: optional string or null`80268027    Replacement system prompt. Up to 100,000 characters. Set to null to clear the agent's system prompt; omit to preserve it.80288029    maxLength: 10000080308031  - `tools: optional array of BetaManagedAgentsAgentToolset20260401Params or BetaManagedAgentsMCPToolsetParams or BetaManagedAgentsCustomToolParams`80328033    Replacement tool list. Full replacement: the provided array becomes the tool configuration. Send an empty array to clear; omit to preserve the agent's tools.80348035    - `BetaManagedAgentsAgentToolset20260401Params object`80368037      Configuration for built-in agent tools. Use this to enable or disable groups of tools available to the agent.80388039      - `type: "agent_toolset_20260401"`80408041      - `configs: optional array of BetaManagedAgentsAgentToolConfigParams`80428043        Per-tool configuration overrides.80448045        - `BetaManagedAgentsBashToolConfigParams object`80468047          Configuration override for the bash tool.80488049          - `type: optional "bash"`80508051          - `name: "bash"`80528053            Must be "bash".80548055          - `enabled: optional boolean or null`80568057            Whether this tool is enabled and available to Claude. Overrides the default_config setting.80588059          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`80608061            Permission policy for tool execution.80628063            - `BetaManagedAgentsAlwaysAllowPolicy object`80648065              Tool calls are automatically approved without user confirmation.80668067              - `type: "always_allow"`80688069            - `BetaManagedAgentsAlwaysAskPolicy object`80708071              Tool calls require user confirmation before execution.80728073              - `type: "always_ask"`80748075            - `BetaManagedAgentsAutoPolicy object`80768077              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.80788079              - `type: "auto"`80808081        - `BetaManagedAgentsEditToolConfigParams object`80828083          Configuration override for the edit tool.80848085          - `type: optional "edit"`80868087          - `name: "edit"`80888089            Must be "edit".80908091          - `enabled: optional boolean or null`80928093            Whether this tool is enabled and available to Claude. Overrides the default_config setting.80948095          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`80968097            Permission policy for tool execution.80988099            - `BetaManagedAgentsAlwaysAllowPolicy object`81008101              Tool calls are automatically approved without user confirmation.81028103            - `BetaManagedAgentsAlwaysAskPolicy object`81048105              Tool calls require user confirmation before execution.81068107            - `BetaManagedAgentsAutoPolicy object`81088109              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.81108111        - `BetaManagedAgentsReadToolConfigParams object`81128113          Configuration override for the read tool.81148115          - `type: optional "read"`81168117          - `name: "read"`81188119            Must be "read".81208121          - `enabled: optional boolean or null`81228123            Whether this tool is enabled and available to Claude. Overrides the default_config setting.81248125          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`81268127            Permission policy for tool execution.81288129            - `BetaManagedAgentsAlwaysAllowPolicy object`81308131              Tool calls are automatically approved without user confirmation.81328133            - `BetaManagedAgentsAlwaysAskPolicy object`81348135              Tool calls require user confirmation before execution.81368137            - `BetaManagedAgentsAutoPolicy object`81388139              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.81408141        - `BetaManagedAgentsWriteToolConfigParams object`81428143          Configuration override for the write tool.81448145          - `type: optional "write"`81468147          - `name: "write"`81488149            Must be "write".81508151          - `enabled: optional boolean or null`81528153            Whether this tool is enabled and available to Claude. Overrides the default_config setting.81548155          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`81568157            Permission policy for tool execution.81588159            - `BetaManagedAgentsAlwaysAllowPolicy object`81608161              Tool calls are automatically approved without user confirmation.81628163            - `BetaManagedAgentsAlwaysAskPolicy object`81648165              Tool calls require user confirmation before execution.81668167            - `BetaManagedAgentsAutoPolicy object`81688169              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.81708171        - `BetaManagedAgentsGlobToolConfigParams object`81728173          Configuration override for the glob tool.81748175          - `type: optional "glob"`81768177          - `name: "glob"`81788179            Must be "glob".81808181          - `enabled: optional boolean or null`81828183            Whether this tool is enabled and available to Claude. Overrides the default_config setting.81848185          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`81868187            Permission policy for tool execution.81888189            - `BetaManagedAgentsAlwaysAllowPolicy object`81908191              Tool calls are automatically approved without user confirmation.81928193            - `BetaManagedAgentsAlwaysAskPolicy object`81948195              Tool calls require user confirmation before execution.81968197            - `BetaManagedAgentsAutoPolicy object`81988199              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.82008201        - `BetaManagedAgentsGrepToolConfigParams object`82028203          Configuration override for the grep tool.82048205          - `type: optional "grep"`82068207          - `name: "grep"`82088209            Must be "grep".82108211          - `enabled: optional boolean or null`82128213            Whether this tool is enabled and available to Claude. Overrides the default_config setting.82148215          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`82168217            Permission policy for tool execution.82188219            - `BetaManagedAgentsAlwaysAllowPolicy object`82208221              Tool calls are automatically approved without user confirmation.82228223            - `BetaManagedAgentsAlwaysAskPolicy object`82248225              Tool calls require user confirmation before execution.82268227            - `BetaManagedAgentsAutoPolicy object`82288229              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.82308231        - `BetaManagedAgentsWebFetchToolConfigParams object`82328233          Configuration override for the web_fetch tool.82348235          - `type: optional "web_fetch"`82368237          - `name: "web_fetch"`82388239            Must be "web_fetch".82408241          - `allowed_domains: optional array of string`82428243            Only fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.82448245          - `blocked_domains: optional array of string`82468247            Never fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.82488249          - `enabled: optional boolean or null`82508251            Whether this tool is enabled and available to Claude. Overrides the default_config setting.82528253          - `max_content_tokens: optional number or null`82548255            Maximum number of tokens of fetched text content to include in context per call. Does not apply to binary content such as PDFs.82568257            format: int3282588259          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`82608261            Permission policy for tool execution.82628263            - `BetaManagedAgentsAlwaysAllowPolicy object`82648265              Tool calls are automatically approved without user confirmation.82668267            - `BetaManagedAgentsAlwaysAskPolicy object`82688269              Tool calls require user confirmation before execution.82708271            - `BetaManagedAgentsAutoPolicy object`82728273              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.82748275        - `BetaManagedAgentsWebSearchToolConfigParams object`82768277          Configuration override for the web_search tool.82788279          - `type: optional "web_search"`82808281          - `name: "web_search"`82828283            Must be "web_search".82848285          - `allowed_domains: optional array of string`82868287            Only return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.82888289          - `blocked_domains: optional array of string`82908291            Never return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.82928293          - `enabled: optional boolean or null`82948295            Whether this tool is enabled and available to Claude. Overrides the default_config setting.82968297          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`82988299            Permission policy for tool execution.83008301            - `BetaManagedAgentsAlwaysAllowPolicy object`83028303              Tool calls are automatically approved without user confirmation.83048305            - `BetaManagedAgentsAlwaysAskPolicy object`83068307              Tool calls require user confirmation before execution.83088309            - `BetaManagedAgentsAutoPolicy object`83108311              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.83128313          - `user_location: optional BetaManagedAgentsUserLocation or null`83148315            Approximate user location for search result localization.83168317            - `type: "approximate"`83188319              Location precision. Only "approximate" is supported.83208321            - `city: optional string or null`83228323              City name.83248325              minLength: 1, maxLength: 25583268327            - `country: optional string or null`83288329              Two-letter ISO 3166-1 country code, uppercase.83308331            - `region: optional string or null`83328333              Region or state name.83348335              minLength: 1, maxLength: 25583368337            - `timezone: optional string or null`83388339              IANA timezone identifier, e.g. "America/Los_Angeles".83408341              minLength: 1, maxLength: 25583428343      - `default_config: optional BetaManagedAgentsAgentToolsetDefaultConfigParams or null`83448345        Default configuration for all tools in a toolset.83468347        - `enabled: optional boolean or null`83488349          Whether tools are enabled and available to Claude by default. Defaults to true if not specified.83508351        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`83528353          Permission policy for tool execution.83548355          - `BetaManagedAgentsAlwaysAllowPolicy object`83568357            Tool calls are automatically approved without user confirmation.83588359          - `BetaManagedAgentsAlwaysAskPolicy object`83608361            Tool calls require user confirmation before execution.83628363          - `BetaManagedAgentsAutoPolicy object`83648365            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.83668367    - `BetaManagedAgentsMCPToolsetParams object`83688369      Configuration for tools from an MCP server defined in `mcp_servers`.83708371      - `type: "mcp_toolset"`83728373      - `mcp_server_name: string`83748375        Name of the MCP server. Must match a server name from the mcp_servers array. 1-255 characters.83768377        minLength: 1, maxLength: 25583788379      - `configs: optional array of BetaManagedAgentsMCPToolConfigParams`83808381        Per-tool configuration overrides.83828383        - `name: string`83848385          Name of the MCP tool to configure. 1-128 characters.83868387          minLength: 1, maxLength: 12883888389        - `enabled: optional boolean or null`83908391          Whether this tool is enabled. Overrides the `default_config` setting.83928393        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`83948395          Permission policy for tool execution.83968397          - `BetaManagedAgentsAlwaysAllowPolicy object`83988399            Tool calls are automatically approved without user confirmation.84008401          - `BetaManagedAgentsAlwaysAskPolicy object`84028403            Tool calls require user confirmation before execution.84048405          - `BetaManagedAgentsAutoPolicy object`84068407            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.84088409      - `default_config: optional BetaManagedAgentsMCPToolsetDefaultConfigParams or null`84108411        Default configuration for all tools from an MCP server.84128413        - `enabled: optional boolean or null`84148415          Whether tools are enabled by default. Defaults to true if not specified.84168417        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`84188419          Permission policy for tool execution.84208421          - `BetaManagedAgentsAlwaysAllowPolicy object`84228423            Tool calls are automatically approved without user confirmation.84248425          - `BetaManagedAgentsAlwaysAskPolicy object`84268427            Tool calls require user confirmation before execution.84288429          - `BetaManagedAgentsAutoPolicy object`84308431            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.84328433    - `BetaManagedAgentsCustomToolParams object`84348435      A custom tool that is executed by the API client rather than the agent. When the agent calls this tool, an `agent.custom_tool_use` event is emitted and the session goes idle, waiting for the client to provide the result via a `user.custom_tool_result` event.84368437      - `type: "custom"`84388439      - `description: string`84408441        Description of what the tool does, shown to the agent to help it decide when to use the tool.84428443        minLength: 184448445      - `input_schema: BetaManagedAgentsCustomToolInputSchema`84468447        JSON Schema for custom tool input parameters.84488449        - `type: "object"`84508451        - `properties: optional map[unknown] or null`84528453        - `required: optional array of string or null`84548455      - `name: string`84568457        Unique name for the tool. 1-128 characters; letters, digits, underscores, and hyphens.84588459        minLength: 1, maxLength: 12884608461  - `version: optional number`84628463    The specific `agent` version to use. Omit to use the latest version.84648465    format: int3284668467### Beta Managed Agents Branch Checkout84688469- `BetaManagedAgentsBranchCheckout object`84708471  - `type: "branch"`84728473  - `name: string`84748475    Branch name to check out.84768477    minLength: 1, maxLength: 25584788479### Beta Managed Agents Budget Limit84808481- `BetaManagedAgentsBudgetLimit object`84828483  A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.84848485  - `type: "limit"`84868487  - `max_list_cost: BetaMonetaryAmount`84888489    A monetary amount in a specific currency.84908491    - `amount: string`84928493      Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.84948495    - `currency: BetaCurrency`84968497      Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.84988499### Beta Managed Agents Cache Creation Usage85008501- `BetaManagedAgentsCacheCreationUsage object`85028503  Prompt-cache creation token usage broken down by cache lifetime.85048505  - `ephemeral_1h_input_tokens: optional number`85068507    Tokens used to create 1-hour ephemeral cache entries.85088509    format: int3285108511  - `ephemeral_5m_input_tokens: optional number`85128513    Tokens used to create 5-minute ephemeral cache entries.85148515    format: int3285168517### Beta Managed Agents Commit Checkout85188519- `BetaManagedAgentsCommitCheckout object`85208521  - `type: "commit"`85228523  - `sha: string`85248525    Full commit SHA to check out.85268527    minLength: 7, maxLength: 6485288529### Beta Managed Agents Deleted Session85308531- `BetaManagedAgentsDeletedSession object`85328533  Confirmation that a `session` has been permanently deleted.85348535  - `type: "session_deleted"`85368537  - `id: string`85388539### Beta Managed Agents Delta Content85408541- `BetaManagedAgentsDeltaContent object`85428543  - `type: "content_delta"`85448545  - `content: BetaManagedAgentsTextBlock`85468547    Regular text content.85488549    - `type: "text"`85508551    - `text: string`85528553      The text content.85548555      minLength: 185568557  - `index: optional number`85588559    Which entry in the previewed event's content array this fragment lands in. Insert content as that entry when the index is new; append to the existing entry otherwise.85608561    format: uint3285628563### Beta Managed Agents Delta Event85648565- `BetaManagedAgentsDeltaEvent object`85668567  An incremental update to an event that is still being streamed. Deltas are best-effort and may stop early; when the buffered event with id == event_id is produced it carries the complete content. A model request that ends early (an error or interrupt) produces no buffered event — its terminal span.model_request_end closes the preview. Only sent on stream connections that opt in via event_deltas; never appears in event history.85688569  - `type: "event_delta"`85708571  - `delta: BetaManagedAgentsDeltaContent`85728573    One fragment of the previewed event. The delta type is named for the previewed event's field it streams into: agent.message events stream content_delta fragments, each a partial element of the content array.85748575    - `type: "content_delta"`85768577    - `content: BetaManagedAgentsTextBlock`85788579      Regular text content.85808581      - `type: "text"`85828583      - `text: string`85848585        The text content.85868587        minLength: 185888589    - `index: optional number`85908591      Which entry in the previewed event's content array this fragment lands in. Insert content as that entry when the index is new; append to the existing entry otherwise.85928593      format: uint3285948595  - `event_id: string`85968597    The id of the event being previewed. Matches event.id on the corresponding event_start and the buffered event that reconciles the preview.85988599### Beta Managed Agents Delta Type86008601- `BetaManagedAgentsDeltaType = "agent.message" or "agent.thinking"`86028603  EventDeltaType enum86048605  - `"agent.message"`86068607  - `"agent.thinking"`86088609### Beta Managed Agents File Resource Params86108611- `BetaManagedAgentsFileResourceParams object`86128613  Mount a file uploaded via the Files API into the session.86148615  - `type: "file"`86168617  - `file_id: string`86188619    ID of a previously uploaded file.86208621    minLength: 1, maxLength: 12886228623  - `mount_path: optional string or null`86248625    Mount path in the container. Defaults to `/mnt/session/uploads/<file_id>`.86268627    minLength: 1, maxLength: 409686288629### Beta Managed Agents GitHub Repository Resource Params86308631- `BetaManagedAgentsGitHubRepositoryResourceParams object`86328633  Mount a GitHub repository into the session's container.86348635  - `type: "github_repository"`86368637  - `url: string`86388639    Github URL of the repository86408641    minLength: 1, maxLength: 204886428643  - `authorization_token: optional string`86448645    GitHub authorization token used to clone the repository. Required for private repositories; optional for public ones.86468647    minLength: 1, maxLength: 409686488649  - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`86508651    Branch or commit to check out. Defaults to the repository's default branch.86528653    - `BetaManagedAgentsBranchCheckout object`86548655      - `type: "branch"`86568657      - `name: string`86588659        Branch name to check out.86608661        minLength: 1, maxLength: 25586628663    - `BetaManagedAgentsCommitCheckout object`86648665      - `type: "commit"`86668667      - `sha: string`86688669        Full commit SHA to check out.86708671        minLength: 7, maxLength: 6486728673  - `mount_path: optional string or null`86748675    Mount path in the container. Defaults to `/workspace/<repo-name>`.86768677    minLength: 1, maxLength: 409686788679### Beta Managed Agents Memory Store Resource Param86808681- `BetaManagedAgentsMemoryStoreResourceParam object`86828683  Parameters for attaching a memory store to an agent session.86848685  - `type: "memory_store"`86868687  - `memory_store_id: string`86888689    The memory store ID (memstore_...). Must belong to the caller's organization and workspace.86908691  - `access: optional "read_write" or "read_only" or null`86928693    Access mode for an attached memory store.86948695    - `"read_write"`86968697    - `"read_only"`86988699  - `instructions: optional string or null`87008701    Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.87028703    maxLength: 409687048705### Beta Managed Agents Multiagent87068707- `BetaManagedAgentsMultiagent object`87088709  Resolved coordinator topology with a concrete agent roster.87108711  - `type: "coordinator"`87128713  - `agents: array of BetaManagedAgentsAgentReference or BetaManagedAgentsAdvisor`87148715    Agents the coordinator may spawn as session threads, each resolved to a specific version.87168717    - `BetaManagedAgentsAgentReference object`87188719      A resolved agent reference with a concrete version.87208721      - `type: "agent"`87228723      - `id: string`87248725      - `version: number`87268727        format: int3287288729    - `BetaManagedAgentsAdvisor object`87308731      Platform advisor roster entry: a model the session's primary thread may consult mid-turn.87328733      - `type: "advisor"`87348735      - `model: string`87368737        The advisor model id.87388739### Beta Managed Agents Multiagent Params87408741- `BetaManagedAgentsMultiagentParams object`87428743  A coordinator topology: the session's primary thread orchestrates work by spawning session threads, each running an agent drawn from the `agents` roster.87448745  - `type: "coordinator"`87468747  - `agents: array of BetaManagedAgentsMultiagentRosterEntryParams`87488749    Agents the coordinator may spawn as session threads. 1–20 entries. Each entry is an agent ID string, a versioned `{"type":"agent","id","version"}` reference, or `{"type":"self"}` to allow recursive self-invocation. Entries must reference distinct agents (after resolving `self` and string forms); at most one `self`. Referenced agents must exist, must not be archived, and must not themselves have `multiagent` set (depth limit 1).87508751    - `string`87528753    - `BetaManagedAgentsAgentParams object`87548755      Specification for an Agent. Provide a specific `version` or use the short-form `agent="agent_id"` for the most recent version87568757      - `type: "agent"`87588759      - `id: string`87608761        The `agent` ID.87628763        minLength: 1, maxLength: 12887648765      - `version: optional number`87668767        The specific `agent` version to use. Omit to use the latest version. Must be at least 1 if specified.87688769        format: int3287708771    - `BetaManagedAgentsMultiagentSelfParams object`87728773      Sentinel roster entry meaning "the agent that owns this configuration". Resolved server-side to a concrete agent reference.87748775      - `type: "self"`87768777    - `BetaManagedAgentsAdvisorParams object`87788779      Platform advisor roster entry: a model the session's primary thread may consult mid-turn. At most one per roster; the entry occupies the roster name `anthropic.advisor`.87808781      - `type: "advisor"`87828783      - `model: string`87848785        A Claude model id. The model must be permitted as an advisor for this agent's model — see the sessions/threads/advisor spec.87868787        minLength: 1, maxLength: 25687888789### Beta Managed Agents Multiagent Roster Entry Params87908791- `BetaManagedAgentsMultiagentRosterEntryParams = string or BetaManagedAgentsAgentParams or BetaManagedAgentsMultiagentSelfParams or BetaManagedAgentsAdvisorParams`87928793  An entry in a multiagent roster: an agent ID string, a versioned agent reference, or `self`.87948795  - `string`87968797  - `BetaManagedAgentsAgentParams object`87988799    Specification for an Agent. Provide a specific `version` or use the short-form `agent="agent_id"` for the most recent version88008801    - `type: "agent"`88028803    - `id: string`88048805      The `agent` ID.88068807      minLength: 1, maxLength: 12888088809    - `version: optional number`88108811      The specific `agent` version to use. Omit to use the latest version. Must be at least 1 if specified.88128813      format: int3288148815  - `BetaManagedAgentsMultiagentSelfParams object`88168817    Sentinel roster entry meaning "the agent that owns this configuration". Resolved server-side to a concrete agent reference.88188819    - `type: "self"`88208821  - `BetaManagedAgentsAdvisorParams object`88228823    Platform advisor roster entry: a model the session's primary thread may consult mid-turn. At most one per roster; the entry occupies the roster name `anthropic.advisor`.88248825    - `type: "advisor"`88268827    - `model: string`88288829      A Claude model id. The model must be permitted as an advisor for this agent's model — see the sessions/threads/advisor spec.88308831      minLength: 1, maxLength: 25688328833### Beta Managed Agents Outcome Evaluation Resource88348835- `BetaManagedAgentsOutcomeEvaluationResource object`88368837  Evaluation state for a single outcome defined via a `define_outcome` event.88388839  - `type: "outcome_evaluation"`88408841  - `completed_at: string or null`88428843    A timestamp in RFC 3339 format88448845    format: date-time88468847  - `description: string`88488849    What the agent should produce.88508851  - `explanation: string or null`88528853    Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.88548855  - `iteration: number`88568857    0-indexed revision cycle the outcome is currently on.88588859    format: int3288608861  - `outcome_id: string`88628863    Server-generated outc_ ID for this outcome.88648865  - `result: string`88668867    Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.88688869### Beta Managed Agents Server Tool Usage88708871- `BetaManagedAgentsServerToolUsage object`88728873  Cumulative count of server-executed tool invocations, broken down by tool.88748875  - `web_fetch_requests: optional number`88768877    Number of server-executed web fetch requests.88788879    format: int3288808881  - `web_search_requests: optional number`88828883    Number of server-executed web search requests.88848885    format: int3288868887### Beta Managed Agents Session88888889- `BetaManagedAgentsSession object`88908891  A Managed Agents `session`.88928893  - `type: "session"`88948895  - `id: string`88968897  - `agent: BetaManagedAgentsSessionAgent`88988899    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.89008901    - `type: "agent"`89028903    - `id: string`89048905    - `description: string or null`89068907    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`89088909      - `type: "url"`89108911      - `name: string`89128913      - `url: string`89148915    - `model: BetaManagedAgentsModelConfig`89168917      Model identifier and configuration.89188919      - `id: BetaManagedAgentsModel`89208921        The model that will power your agent.89228923        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.89248925        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`89268927          The model that will power your agent.89288929          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.89308931          - `"claude-fable-5-1"`89328933            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows89348935          - `"claude-sonnet-5"`89368937            High-performance model for coding and agents89388939          - `"claude-fable-5"`89408941            Next generation of intelligence for the hardest knowledge work and coding problems89428943          - `"claude-opus-5"`89448945            Powerful intelligence for long-running agents and coding89468947          - `"claude-opus-4-8"`89488949            Powerful intelligence for long-running agents and coding89508951          - `"claude-opus-4-7"`89528953            Powerful intelligence for long-running agents and coding89548955          - `"claude-opus-4-6"`89568957            Powerful intelligence for long-running agents and coding89588959          - `"claude-sonnet-4-6"`89608961            Best combination of speed and intelligence89628963          - `"claude-haiku-4-5"`89648965            Fastest model with near-frontier intelligence89668967          - `"claude-haiku-4-5-20251001"`89688969            Fastest model with near-frontier intelligence89708971          - `"claude-opus-4-5"`89728973            Powerful intelligence for long-running agents and coding89748975          - `"claude-opus-4-5-20251101"`89768977            Powerful intelligence for long-running agents and coding89788979          - `"claude-sonnet-4-5"`89808981            High-performance model for agents and coding89828983          - `"claude-sonnet-4-5-20250929"`89848985            High-performance model for agents and coding89868987        - `string`89888989      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`89908991        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.89928993        - `BetaManagedAgentsEffortLow object`89948995          Low effort. Favors latency over reasoning depth.89968997          - `type: "low"`89988999        - `BetaManagedAgentsEffortMedium object`90009001          Medium effort. Balances latency and reasoning depth.90029003          - `type: "medium"`90049005        - `BetaManagedAgentsEffortHigh object`90069007          High effort. Favors reasoning depth.90089009          - `type: "high"`90109011        - `BetaManagedAgentsEffortXhigh object`90129013          Extra-high effort. Not all models accept this level.90149015          - `type: "xhigh"`90169017        - `BetaManagedAgentsEffortMax object`90189019          Maximum effort. Favors reasoning depth over latency.90209021          - `type: "max"`90229023      - `inference_geo: optional string`90249025        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.90269027      - `speed: optional "standard" or "fast"`90289029        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.90309031        - `"standard"`90329033        - `"fast"`90349035    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`90369037      Resolved coordinator topology with full agent definitions for each roster member.90389039      - `type: "coordinator"`90409041      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`90429043        Full `agent` definitions the coordinator may spawn as session threads.90449045        - `BetaManagedAgentsSessionThreadAgent object`90469047          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.90489049          - `type: "agent"`90509051          - `id: string`90529053          - `description: string or null`90549055          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`90569057            - `type: "url"`90589059            - `name: string`90609061            - `url: string`90629063          - `model: BetaManagedAgentsModelConfig`90649065            Model identifier and configuration.90669067          - `name: string`90689069          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`90709071            - `BetaManagedAgentsAnthropicSkill object`90729073              A resolved Anthropic-managed skill.90749075              - `type: "anthropic"`90769077              - `skill_id: string`90789079              - `version: string`90809081            - `BetaManagedAgentsCustomSkill object`90829083              A resolved user-created custom skill.90849085              - `type: "custom"`90869087              - `skill_id: string`90889089              - `version: string`90909091          - `system: string or null`90929093          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`90949095            - `BetaManagedAgentsAgentToolset20260401 object`90969097              - `type: "agent_toolset_20260401"`90989099              - `configs: array of BetaManagedAgentsAgentToolConfig`91009101                - `BetaManagedAgentsBashToolConfig object`91029103                  Configuration for the bash tool.91049105                  - `type: "bash"`91069107                  - `enabled: boolean`91089109                  - `name: "bash"`91109111                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`91129113                    Permission policy for tool execution.91149115                    - `BetaManagedAgentsAlwaysAllowPolicy object`91169117                      Tool calls are automatically approved without user confirmation.91189119                      - `type: "always_allow"`91209121                    - `BetaManagedAgentsAlwaysAskPolicy object`91229123                      Tool calls require user confirmation before execution.91249125                      - `type: "always_ask"`91269127                    - `BetaManagedAgentsAutoPolicy object`91289129                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.91309131                      - `type: "auto"`91329133                - `BetaManagedAgentsEditToolConfig object`91349135                  Configuration for the edit tool.91369137                  - `type: "edit"`91389139                  - `enabled: boolean`91409141                  - `name: "edit"`91429143                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`91449145                    Permission policy for tool execution.91469147                    - `BetaManagedAgentsAlwaysAllowPolicy object`91489149                      Tool calls are automatically approved without user confirmation.91509151                    - `BetaManagedAgentsAlwaysAskPolicy object`91529153                      Tool calls require user confirmation before execution.91549155                    - `BetaManagedAgentsAutoPolicy object`91569157                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.91589159                - `BetaManagedAgentsReadToolConfig object`91609161                  Configuration for the read tool.91629163                  - `type: "read"`91649165                  - `enabled: boolean`91669167                  - `name: "read"`91689169                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`91709171                    Permission policy for tool execution.91729173                    - `BetaManagedAgentsAlwaysAllowPolicy object`91749175                      Tool calls are automatically approved without user confirmation.91769177                    - `BetaManagedAgentsAlwaysAskPolicy object`91789179                      Tool calls require user confirmation before execution.91809181                    - `BetaManagedAgentsAutoPolicy object`91829183                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.91849185                - `BetaManagedAgentsWriteToolConfig object`91869187                  Configuration for the write tool.91889189                  - `type: "write"`91909191                  - `enabled: boolean`91929193                  - `name: "write"`91949195                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`91969197                    Permission policy for tool execution.91989199                    - `BetaManagedAgentsAlwaysAllowPolicy object`92009201                      Tool calls are automatically approved without user confirmation.92029203                    - `BetaManagedAgentsAlwaysAskPolicy object`92049205                      Tool calls require user confirmation before execution.92069207                    - `BetaManagedAgentsAutoPolicy object`92089209                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.92109211                - `BetaManagedAgentsGlobToolConfig object`92129213                  Configuration for the glob tool.92149215                  - `type: "glob"`92169217                  - `enabled: boolean`92189219                  - `name: "glob"`92209221                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`92229223                    Permission policy for tool execution.92249225                    - `BetaManagedAgentsAlwaysAllowPolicy object`92269227                      Tool calls are automatically approved without user confirmation.92289229                    - `BetaManagedAgentsAlwaysAskPolicy object`92309231                      Tool calls require user confirmation before execution.92329233                    - `BetaManagedAgentsAutoPolicy object`92349235                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.92369237                - `BetaManagedAgentsGrepToolConfig object`92389239                  Configuration for the grep tool.92409241                  - `type: "grep"`92429243                  - `enabled: boolean`92449245                  - `name: "grep"`92469247                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`92489249                    Permission policy for tool execution.92509251                    - `BetaManagedAgentsAlwaysAllowPolicy object`92529253                      Tool calls are automatically approved without user confirmation.92549255                    - `BetaManagedAgentsAlwaysAskPolicy object`92569257                      Tool calls require user confirmation before execution.92589259                    - `BetaManagedAgentsAutoPolicy object`92609261                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.92629263                - `BetaManagedAgentsWebFetchToolConfig object`92649265                  Configuration for the web_fetch tool.92669267                  - `type: "web_fetch"`92689269                  - `enabled: boolean`92709271                  - `name: "web_fetch"`92729273                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`92749275                    Permission policy for tool execution.92769277                    - `BetaManagedAgentsAlwaysAllowPolicy object`92789279                      Tool calls are automatically approved without user confirmation.92809281                    - `BetaManagedAgentsAlwaysAskPolicy object`92829283                      Tool calls require user confirmation before execution.92849285                    - `BetaManagedAgentsAutoPolicy object`92869287                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.92889289                  - `allowed_domains: optional array of string`92909291                  - `blocked_domains: optional array of string`92929293                  - `max_content_tokens: optional number or null`92949295                    format: int3292969297                - `BetaManagedAgentsWebSearchToolConfig object`92989299                  Configuration for the web_search tool.93009301                  - `type: "web_search"`93029303                  - `enabled: boolean`93049305                  - `name: "web_search"`93069307                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`93089309                    Permission policy for tool execution.93109311                    - `BetaManagedAgentsAlwaysAllowPolicy object`93129313                      Tool calls are automatically approved without user confirmation.93149315                    - `BetaManagedAgentsAlwaysAskPolicy object`93169317                      Tool calls require user confirmation before execution.93189319                    - `BetaManagedAgentsAutoPolicy object`93209321                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.93229323                  - `allowed_domains: optional array of string`93249325                  - `blocked_domains: optional array of string`93269327                  - `user_location: optional BetaManagedAgentsUserLocation or null`93289329                    Approximate user location for search result localization.93309331                    - `type: "approximate"`93329333                      Location precision. Only "approximate" is supported.93349335                    - `city: optional string or null`93369337                      City name.93389339                      minLength: 1, maxLength: 25593409341                    - `country: optional string or null`93429343                      Two-letter ISO 3166-1 country code, uppercase.93449345                    - `region: optional string or null`93469347                      Region or state name.93489349                      minLength: 1, maxLength: 25593509351                    - `timezone: optional string or null`93529353                      IANA timezone identifier, e.g. "America/Los_Angeles".93549355                      minLength: 1, maxLength: 25593569357              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`93589359                Resolved default configuration for agent tools.93609361                - `enabled: boolean`93629363                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`93649365                  Permission policy for tool execution.93669367                  - `BetaManagedAgentsAlwaysAllowPolicy object`93689369                    Tool calls are automatically approved without user confirmation.93709371                  - `BetaManagedAgentsAlwaysAskPolicy object`93729373                    Tool calls require user confirmation before execution.93749375                  - `BetaManagedAgentsAutoPolicy object`93769377                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.93789379            - `BetaManagedAgentsMCPToolset object`93809381              - `type: "mcp_toolset"`93829383              - `configs: array of BetaManagedAgentsMCPToolConfig`93849385                - `enabled: boolean`93869387                - `name: string`93889389                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`93909391                  Permission policy for tool execution.93929393                  - `BetaManagedAgentsAlwaysAllowPolicy object`93949395                    Tool calls are automatically approved without user confirmation.93969397                  - `BetaManagedAgentsAlwaysAskPolicy object`93989399                    Tool calls require user confirmation before execution.94009401                  - `BetaManagedAgentsAutoPolicy object`94029403                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.94049405              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`94069407                Resolved default configuration for all tools from an MCP server.94089409                - `enabled: boolean`94109411                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`94129413                  Permission policy for tool execution.94149415                  - `BetaManagedAgentsAlwaysAllowPolicy object`94169417                    Tool calls are automatically approved without user confirmation.94189419                  - `BetaManagedAgentsAlwaysAskPolicy object`94209421                    Tool calls require user confirmation before execution.94229423                  - `BetaManagedAgentsAutoPolicy object`94249425                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.94269427              - `mcp_server_name: string`94289429            - `BetaManagedAgentsCustomTool object`94309431              A custom tool as returned in API responses.94329433              - `type: "custom"`94349435              - `description: string`94369437              - `input_schema: BetaManagedAgentsCustomToolInputSchema`94389439                JSON Schema for custom tool input parameters.94409441                - `type: "object"`94429443                - `properties: optional map[unknown] or null`94449445                - `required: optional array of string or null`94469447              - `name: string`94489449          - `version: number`94509451            format: int3294529453        - `BetaManagedAgentsAdvisor object`94549455          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.94569457          - `type: "advisor"`94589459          - `model: string`94609461            The advisor model id.94629463    - `name: string`94649465    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`94669467      - `BetaManagedAgentsAnthropicSkill object`94689469        A resolved Anthropic-managed skill.94709471      - `BetaManagedAgentsCustomSkill object`94729473        A resolved user-created custom skill.94749475    - `system: string or null`94769477    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`94789479      - `BetaManagedAgentsAgentToolset20260401 object`94809481      - `BetaManagedAgentsMCPToolset object`94829483      - `BetaManagedAgentsCustomTool object`94849485        A custom tool as returned in API responses.94869487    - `version: number`94889489      format: int3294909491  - `archived_at: string or null`94929493    A timestamp in RFC 3339 format94949495    format: date-time94969497  - `budget: BetaManagedAgentsBudgetLimit or null`94989499    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.95009501    - `type: "limit"`95029503    - `max_list_cost: BetaMonetaryAmount`95049505      A monetary amount in a specific currency.95069507      - `amount: string`95089509        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.95109511      - `currency: BetaCurrency`95129513        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.95149515  - `created_at: string`95169517    A timestamp in RFC 3339 format95189519    format: date-time95209521  - `environment_id: string`95229523  - `metadata: map[string]`95249525  - `outcome_evaluations: array of BetaManagedAgentsOutcomeEvaluationResource`95269527    Per-outcome evaluation state. One entry per `define_outcome` event sent to the session.95289529    - `type: "outcome_evaluation"`95309531    - `completed_at: string or null`95329533      A timestamp in RFC 3339 format95349535      format: date-time95369537    - `description: string`95389539      What the agent should produce.95409541    - `explanation: string or null`95429543      Grader's verdict text from the most recent evaluation. For `satisfied`, explains why criteria are met; for `needs_revision` (intermediate), what's missing; for `failed`, why unrecoverable.95449545    - `iteration: number`95469547      0-indexed revision cycle the outcome is currently on.95489549      format: int3295509551    - `outcome_id: string`95529553      Server-generated outc_ ID for this outcome.95549555    - `result: string`95569557      Current evaluation state. `pending` before the agent begins work; `running` while producing or revising; `evaluating` while the grader scores; `satisfied`/`max_iterations_reached`/`failed`/`interrupted` are terminal.95589559  - `resources: array of BetaManagedAgentsSessionResource`95609561    - `BetaManagedAgentsGitHubRepositoryResource object`95629563      - `type: "github_repository"`95649565      - `id: string`95669567      - `created_at: string`95689569        A timestamp in RFC 3339 format95709571        format: date-time95729573      - `mount_path: string`95749575      - `updated_at: string`95769577        A timestamp in RFC 3339 format95789579        format: date-time95809581      - `url: string`95829583      - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`95849585        - `BetaManagedAgentsBranchCheckout object`95869587          - `type: "branch"`95889589          - `name: string`95909591            Branch name to check out.95929593            minLength: 1, maxLength: 25595949595        - `BetaManagedAgentsCommitCheckout object`95969597          - `type: "commit"`95989599          - `sha: string`96009601            Full commit SHA to check out.96029603            minLength: 7, maxLength: 6496049605    - `BetaManagedAgentsFileResource object`96069607      - `type: "file"`96089609      - `id: string`96109611      - `created_at: string`96129613        A timestamp in RFC 3339 format96149615        format: date-time96169617      - `file_id: string`96189619      - `mount_path: string`96209621      - `updated_at: string`96229623        A timestamp in RFC 3339 format96249625        format: date-time96269627    - `BetaManagedAgentsMemoryStoreResource object`96289629      A memory store attached to an agent session.96309631      - `type: "memory_store"`96329633      - `memory_store_id: string`96349635        The memory store ID (memstore_...). Must belong to the caller's organization and workspace.96369637      - `access: optional "read_write" or "read_only" or null`96389639        Access mode for an attached memory store.96409641        - `"read_write"`96429643        - `"read_only"`96449645      - `description: optional string`96469647        Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.96489649      - `instructions: optional string or null`96509651        Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.96529653        maxLength: 409696549655      - `mount_path: optional string or null`96569657        Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.96589659      - `name: optional string or null`96609661        Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.96629663  - `stats: BetaManagedAgentsSessionStats`96649665    Timing statistics for a session.96669667    - `active_seconds: optional number`96689669      Cumulative time in seconds the session spent in `running` status. Excludes idle time.96709671      format: double96729673    - `duration_seconds: optional number`96749675      Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.96769677      format: double96789679  - `status: "rescheduling" or "running" or "idle" or "terminated"`96809681    SessionStatus enum96829683    - `"rescheduling"`96849685    - `"running"`96869687    - `"idle"`96889689    - `"terminated"`96909691  - `title: string or null`96929693  - `updated_at: string`96949695    A timestamp in RFC 3339 format96969697    format: date-time96989699  - `usage: BetaManagedAgentsSessionUsage`97009701    Cumulative token usage for a session across all turns.97029703    - `active_seconds: optional number`97049705      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.97069707      format: double97089709    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`97109711      Prompt-cache creation token usage broken down by cache lifetime.97129713      - `ephemeral_1h_input_tokens: optional number`97149715        Tokens used to create 1-hour ephemeral cache entries.97169717        format: int3297189719      - `ephemeral_5m_input_tokens: optional number`97209721        Tokens used to create 5-minute ephemeral cache entries.97229723        format: int3297249725    - `cache_read_input_tokens: optional number`97269727      Total tokens read from prompt cache.97289729      format: int3297309731    - `input_tokens: optional number`97329733      Total input tokens consumed across all turns.97349735      format: int3297369737    - `list_cost: optional BetaMonetaryAmount or null`97389739      A monetary amount in a specific currency.97409741    - `output_tokens: optional number`97429743      Total output tokens generated across all turns.97449745      format: int3297469747    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`97489749      Cumulative count of server-executed tool invocations, broken down by tool.97509751      - `web_fetch_requests: optional number`97529753        Number of server-executed web fetch requests.97549755        format: int3297569757      - `web_search_requests: optional number`97589759        Number of server-executed web search requests.97609761        format: int3297629763  - `vault_ids: array of string`97649765    Vault IDs attached to the session at creation. Empty when no vaults were supplied.97669767  - `deployment_id: optional string or null`97689769    Deployment ID when the session was created from a deployment reference. Null otherwise.97709771### Beta Managed Agents Session Agent97729773- `BetaManagedAgentsSessionAgent object`97749775  Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.97769777  - `type: "agent"`97789779  - `id: string`97809781  - `description: string or null`97829783  - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`97849785    - `type: "url"`97869787    - `name: string`97889789    - `url: string`97909791  - `model: BetaManagedAgentsModelConfig`97929793    Model identifier and configuration.97949795    - `id: BetaManagedAgentsModel`97969797      The model that will power your agent.97989799      See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.98009801      - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`98029803        The model that will power your agent.98049805        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.98069807        - `"claude-fable-5-1"`98089809          Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows98109811        - `"claude-sonnet-5"`98129813          High-performance model for coding and agents98149815        - `"claude-fable-5"`98169817          Next generation of intelligence for the hardest knowledge work and coding problems98189819        - `"claude-opus-5"`98209821          Powerful intelligence for long-running agents and coding98229823        - `"claude-opus-4-8"`98249825          Powerful intelligence for long-running agents and coding98269827        - `"claude-opus-4-7"`98289829          Powerful intelligence for long-running agents and coding98309831        - `"claude-opus-4-6"`98329833          Powerful intelligence for long-running agents and coding98349835        - `"claude-sonnet-4-6"`98369837          Best combination of speed and intelligence98389839        - `"claude-haiku-4-5"`98409841          Fastest model with near-frontier intelligence98429843        - `"claude-haiku-4-5-20251001"`98449845          Fastest model with near-frontier intelligence98469847        - `"claude-opus-4-5"`98489849          Powerful intelligence for long-running agents and coding98509851        - `"claude-opus-4-5-20251101"`98529853          Powerful intelligence for long-running agents and coding98549855        - `"claude-sonnet-4-5"`98569857          High-performance model for agents and coding98589859        - `"claude-sonnet-4-5-20250929"`98609861          High-performance model for agents and coding98629863      - `string`98649865    - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`98669867      How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.98689869      - `BetaManagedAgentsEffortLow object`98709871        Low effort. Favors latency over reasoning depth.98729873        - `type: "low"`98749875      - `BetaManagedAgentsEffortMedium object`98769877        Medium effort. Balances latency and reasoning depth.98789879        - `type: "medium"`98809881      - `BetaManagedAgentsEffortHigh object`98829883        High effort. Favors reasoning depth.98849885        - `type: "high"`98869887      - `BetaManagedAgentsEffortXhigh object`98889889        Extra-high effort. Not all models accept this level.98909891        - `type: "xhigh"`98929893      - `BetaManagedAgentsEffortMax object`98949895        Maximum effort. Favors reasoning depth over latency.98969897        - `type: "max"`98989899    - `inference_geo: optional string`99009901      Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.99029903    - `speed: optional "standard" or "fast"`99049905      Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.99069907      - `"standard"`99089909      - `"fast"`99109911  - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`99129913    Resolved coordinator topology with full agent definitions for each roster member.99149915    - `type: "coordinator"`99169917    - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`99189919      Full `agent` definitions the coordinator may spawn as session threads.99209921      - `BetaManagedAgentsSessionThreadAgent object`99229923        Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.99249925        - `type: "agent"`99269927        - `id: string`99289929        - `description: string or null`99309931        - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`99329933          - `type: "url"`99349935          - `name: string`99369937          - `url: string`99389939        - `model: BetaManagedAgentsModelConfig`99409941          Model identifier and configuration.99429943        - `name: string`99449945        - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`99469947          - `BetaManagedAgentsAnthropicSkill object`99489949            A resolved Anthropic-managed skill.99509951            - `type: "anthropic"`99529953            - `skill_id: string`99549955            - `version: string`99569957          - `BetaManagedAgentsCustomSkill object`99589959            A resolved user-created custom skill.99609961            - `type: "custom"`99629963            - `skill_id: string`99649965            - `version: string`99669967        - `system: string or null`99689969        - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`99709971          - `BetaManagedAgentsAgentToolset20260401 object`99729973            - `type: "agent_toolset_20260401"`99749975            - `configs: array of BetaManagedAgentsAgentToolConfig`99769977              - `BetaManagedAgentsBashToolConfig object`99789979                Configuration for the bash tool.99809981                - `type: "bash"`99829983                - `enabled: boolean`99849985                - `name: "bash"`99869987                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`99889989                  Permission policy for tool execution.99909991                  - `BetaManagedAgentsAlwaysAllowPolicy object`99929993                    Tool calls are automatically approved without user confirmation.99949995                    - `type: "always_allow"`99969997                  - `BetaManagedAgentsAlwaysAskPolicy object`99989999                    Tool calls require user confirmation before execution.1000010001                    - `type: "always_ask"`1000210003                  - `BetaManagedAgentsAutoPolicy object`1000410005                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1000610007                    - `type: "auto"`1000810009              - `BetaManagedAgentsEditToolConfig object`1001010011                Configuration for the edit tool.1001210013                - `type: "edit"`1001410015                - `enabled: boolean`1001610017                - `name: "edit"`1001810019                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1002010021                  Permission policy for tool execution.1002210023                  - `BetaManagedAgentsAlwaysAllowPolicy object`1002410025                    Tool calls are automatically approved without user confirmation.1002610027                  - `BetaManagedAgentsAlwaysAskPolicy object`1002810029                    Tool calls require user confirmation before execution.1003010031                  - `BetaManagedAgentsAutoPolicy object`1003210033                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1003410035              - `BetaManagedAgentsReadToolConfig object`1003610037                Configuration for the read tool.1003810039                - `type: "read"`1004010041                - `enabled: boolean`1004210043                - `name: "read"`1004410045                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1004610047                  Permission policy for tool execution.1004810049                  - `BetaManagedAgentsAlwaysAllowPolicy object`1005010051                    Tool calls are automatically approved without user confirmation.1005210053                  - `BetaManagedAgentsAlwaysAskPolicy object`1005410055                    Tool calls require user confirmation before execution.1005610057                  - `BetaManagedAgentsAutoPolicy object`1005810059                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1006010061              - `BetaManagedAgentsWriteToolConfig object`1006210063                Configuration for the write tool.1006410065                - `type: "write"`1006610067                - `enabled: boolean`1006810069                - `name: "write"`1007010071                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1007210073                  Permission policy for tool execution.1007410075                  - `BetaManagedAgentsAlwaysAllowPolicy object`1007610077                    Tool calls are automatically approved without user confirmation.1007810079                  - `BetaManagedAgentsAlwaysAskPolicy object`1008010081                    Tool calls require user confirmation before execution.1008210083                  - `BetaManagedAgentsAutoPolicy object`1008410085                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1008610087              - `BetaManagedAgentsGlobToolConfig object`1008810089                Configuration for the glob tool.1009010091                - `type: "glob"`1009210093                - `enabled: boolean`1009410095                - `name: "glob"`1009610097                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1009810099                  Permission policy for tool execution.1010010101                  - `BetaManagedAgentsAlwaysAllowPolicy object`1010210103                    Tool calls are automatically approved without user confirmation.1010410105                  - `BetaManagedAgentsAlwaysAskPolicy object`1010610107                    Tool calls require user confirmation before execution.1010810109                  - `BetaManagedAgentsAutoPolicy object`1011010111                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1011210113              - `BetaManagedAgentsGrepToolConfig object`1011410115                Configuration for the grep tool.1011610117                - `type: "grep"`1011810119                - `enabled: boolean`1012010121                - `name: "grep"`1012210123                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1012410125                  Permission policy for tool execution.1012610127                  - `BetaManagedAgentsAlwaysAllowPolicy object`1012810129                    Tool calls are automatically approved without user confirmation.1013010131                  - `BetaManagedAgentsAlwaysAskPolicy object`1013210133                    Tool calls require user confirmation before execution.1013410135                  - `BetaManagedAgentsAutoPolicy object`1013610137                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1013810139              - `BetaManagedAgentsWebFetchToolConfig object`1014010141                Configuration for the web_fetch tool.1014210143                - `type: "web_fetch"`1014410145                - `enabled: boolean`1014610147                - `name: "web_fetch"`1014810149                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1015010151                  Permission policy for tool execution.1015210153                  - `BetaManagedAgentsAlwaysAllowPolicy object`1015410155                    Tool calls are automatically approved without user confirmation.1015610157                  - `BetaManagedAgentsAlwaysAskPolicy object`1015810159                    Tool calls require user confirmation before execution.1016010161                  - `BetaManagedAgentsAutoPolicy object`1016210163                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1016410165                - `allowed_domains: optional array of string`1016610167                - `blocked_domains: optional array of string`1016810169                - `max_content_tokens: optional number or null`1017010171                  format: int321017210173              - `BetaManagedAgentsWebSearchToolConfig object`1017410175                Configuration for the web_search tool.1017610177                - `type: "web_search"`1017810179                - `enabled: boolean`1018010181                - `name: "web_search"`1018210183                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1018410185                  Permission policy for tool execution.1018610187                  - `BetaManagedAgentsAlwaysAllowPolicy object`1018810189                    Tool calls are automatically approved without user confirmation.1019010191                  - `BetaManagedAgentsAlwaysAskPolicy object`1019210193                    Tool calls require user confirmation before execution.1019410195                  - `BetaManagedAgentsAutoPolicy object`1019610197                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1019810199                - `allowed_domains: optional array of string`1020010201                - `blocked_domains: optional array of string`1020210203                - `user_location: optional BetaManagedAgentsUserLocation or null`1020410205                  Approximate user location for search result localization.1020610207                  - `type: "approximate"`1020810209                    Location precision. Only "approximate" is supported.1021010211                  - `city: optional string or null`1021210213                    City name.1021410215                    minLength: 1, maxLength: 2551021610217                  - `country: optional string or null`1021810219                    Two-letter ISO 3166-1 country code, uppercase.1022010221                  - `region: optional string or null`1022210223                    Region or state name.1022410225                    minLength: 1, maxLength: 2551022610227                  - `timezone: optional string or null`1022810229                    IANA timezone identifier, e.g. "America/Los_Angeles".1023010231                    minLength: 1, maxLength: 2551023210233            - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`1023410235              Resolved default configuration for agent tools.1023610237              - `enabled: boolean`1023810239              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1024010241                Permission policy for tool execution.1024210243                - `BetaManagedAgentsAlwaysAllowPolicy object`1024410245                  Tool calls are automatically approved without user confirmation.1024610247                - `BetaManagedAgentsAlwaysAskPolicy object`1024810249                  Tool calls require user confirmation before execution.1025010251                - `BetaManagedAgentsAutoPolicy object`1025210253                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1025410255          - `BetaManagedAgentsMCPToolset object`1025610257            - `type: "mcp_toolset"`1025810259            - `configs: array of BetaManagedAgentsMCPToolConfig`1026010261              - `enabled: boolean`1026210263              - `name: string`1026410265              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1026610267                Permission policy for tool execution.1026810269                - `BetaManagedAgentsAlwaysAllowPolicy object`1027010271                  Tool calls are automatically approved without user confirmation.1027210273                - `BetaManagedAgentsAlwaysAskPolicy object`1027410275                  Tool calls require user confirmation before execution.1027610277                - `BetaManagedAgentsAutoPolicy object`1027810279                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1028010281            - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`1028210283              Resolved default configuration for all tools from an MCP server.1028410285              - `enabled: boolean`1028610287              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1028810289                Permission policy for tool execution.1029010291                - `BetaManagedAgentsAlwaysAllowPolicy object`1029210293                  Tool calls are automatically approved without user confirmation.1029410295                - `BetaManagedAgentsAlwaysAskPolicy object`1029610297                  Tool calls require user confirmation before execution.1029810299                - `BetaManagedAgentsAutoPolicy object`1030010301                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1030210303            - `mcp_server_name: string`1030410305          - `BetaManagedAgentsCustomTool object`1030610307            A custom tool as returned in API responses.1030810309            - `type: "custom"`1031010311            - `description: string`1031210313            - `input_schema: BetaManagedAgentsCustomToolInputSchema`1031410315              JSON Schema for custom tool input parameters.1031610317              - `type: "object"`1031810319              - `properties: optional map[unknown] or null`1032010321              - `required: optional array of string or null`1032210323            - `name: string`1032410325        - `version: number`1032610327          format: int321032810329      - `BetaManagedAgentsAdvisor object`1033010331        Platform advisor roster entry: a model the session's primary thread may consult mid-turn.1033210333        - `type: "advisor"`1033410335        - `model: string`1033610337          The advisor model id.1033810339  - `name: string`1034010341  - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1034210343    - `BetaManagedAgentsAnthropicSkill object`1034410345      A resolved Anthropic-managed skill.1034610347    - `BetaManagedAgentsCustomSkill object`1034810349      A resolved user-created custom skill.1035010351  - `system: string or null`1035210353  - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1035410355    - `BetaManagedAgentsAgentToolset20260401 object`1035610357    - `BetaManagedAgentsMCPToolset object`1035810359    - `BetaManagedAgentsCustomTool object`1036010361      A custom tool as returned in API responses.1036210363  - `version: number`1036410365    format: int321036610367### Beta Managed Agents Session Agent Update1036810369- `BetaManagedAgentsSessionAgentUpdate object`1037010371  Mid-session agent configuration update. Only `tools` and `mcp_servers` are updatable. Full replacement: the provided array becomes the new value. To preserve existing entries, GET the session, modify the array, and POST it back.1037210373  - `mcp_servers: optional array of BetaManagedAgentsURLMCPServerParams`1037410375    Replacement MCP server list. Full replacement: the provided array becomes the new value. Send an empty array to clear; omit to preserve.1037610377    - `type: "url"`1037810379    - `name: string`1038010381      Unique name for this server, referenced by mcp_toolset configurations. 1-255 characters.1038210383      minLength: 1, maxLength: 2551038410385    - `url: string`1038610387      Endpoint URL for the MCP server.1038810389      maxLength: 20481039010391  - `tools: optional array of BetaManagedAgentsAgentToolset20260401Params or BetaManagedAgentsMCPToolsetParams or BetaManagedAgentsCustomToolParams`1039210393    Replacement tool list. Full replacement: the provided array becomes the new value. Send an empty array to clear; omit to preserve.1039410395    - `BetaManagedAgentsAgentToolset20260401Params object`1039610397      Configuration for built-in agent tools. Use this to enable or disable groups of tools available to the agent.1039810399      - `type: "agent_toolset_20260401"`1040010401      - `configs: optional array of BetaManagedAgentsAgentToolConfigParams`1040210403        Per-tool configuration overrides.1040410405        - `BetaManagedAgentsBashToolConfigParams object`1040610407          Configuration override for the bash tool.1040810409          - `type: optional "bash"`1041010411          - `name: "bash"`1041210413            Must be "bash".1041410415          - `enabled: optional boolean or null`1041610417            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1041810419          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1042010421            Permission policy for tool execution.1042210423            - `BetaManagedAgentsAlwaysAllowPolicy object`1042410425              Tool calls are automatically approved without user confirmation.1042610427              - `type: "always_allow"`1042810429            - `BetaManagedAgentsAlwaysAskPolicy object`1043010431              Tool calls require user confirmation before execution.1043210433              - `type: "always_ask"`1043410435            - `BetaManagedAgentsAutoPolicy object`1043610437              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1043810439              - `type: "auto"`1044010441        - `BetaManagedAgentsEditToolConfigParams object`1044210443          Configuration override for the edit tool.1044410445          - `type: optional "edit"`1044610447          - `name: "edit"`1044810449            Must be "edit".1045010451          - `enabled: optional boolean or null`1045210453            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1045410455          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1045610457            Permission policy for tool execution.1045810459            - `BetaManagedAgentsAlwaysAllowPolicy object`1046010461              Tool calls are automatically approved without user confirmation.1046210463            - `BetaManagedAgentsAlwaysAskPolicy object`1046410465              Tool calls require user confirmation before execution.1046610467            - `BetaManagedAgentsAutoPolicy object`1046810469              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1047010471        - `BetaManagedAgentsReadToolConfigParams object`1047210473          Configuration override for the read tool.1047410475          - `type: optional "read"`1047610477          - `name: "read"`1047810479            Must be "read".1048010481          - `enabled: optional boolean or null`1048210483            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1048410485          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1048610487            Permission policy for tool execution.1048810489            - `BetaManagedAgentsAlwaysAllowPolicy object`1049010491              Tool calls are automatically approved without user confirmation.1049210493            - `BetaManagedAgentsAlwaysAskPolicy object`1049410495              Tool calls require user confirmation before execution.1049610497            - `BetaManagedAgentsAutoPolicy object`1049810499              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1050010501        - `BetaManagedAgentsWriteToolConfigParams object`1050210503          Configuration override for the write tool.1050410505          - `type: optional "write"`1050610507          - `name: "write"`1050810509            Must be "write".1051010511          - `enabled: optional boolean or null`1051210513            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1051410515          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1051610517            Permission policy for tool execution.1051810519            - `BetaManagedAgentsAlwaysAllowPolicy object`1052010521              Tool calls are automatically approved without user confirmation.1052210523            - `BetaManagedAgentsAlwaysAskPolicy object`1052410525              Tool calls require user confirmation before execution.1052610527            - `BetaManagedAgentsAutoPolicy object`1052810529              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1053010531        - `BetaManagedAgentsGlobToolConfigParams object`1053210533          Configuration override for the glob tool.1053410535          - `type: optional "glob"`1053610537          - `name: "glob"`1053810539            Must be "glob".1054010541          - `enabled: optional boolean or null`1054210543            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1054410545          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1054610547            Permission policy for tool execution.1054810549            - `BetaManagedAgentsAlwaysAllowPolicy object`1055010551              Tool calls are automatically approved without user confirmation.1055210553            - `BetaManagedAgentsAlwaysAskPolicy object`1055410555              Tool calls require user confirmation before execution.1055610557            - `BetaManagedAgentsAutoPolicy object`1055810559              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1056010561        - `BetaManagedAgentsGrepToolConfigParams object`1056210563          Configuration override for the grep tool.1056410565          - `type: optional "grep"`1056610567          - `name: "grep"`1056810569            Must be "grep".1057010571          - `enabled: optional boolean or null`1057210573            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1057410575          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1057610577            Permission policy for tool execution.1057810579            - `BetaManagedAgentsAlwaysAllowPolicy object`1058010581              Tool calls are automatically approved without user confirmation.1058210583            - `BetaManagedAgentsAlwaysAskPolicy object`1058410585              Tool calls require user confirmation before execution.1058610587            - `BetaManagedAgentsAutoPolicy object`1058810589              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1059010591        - `BetaManagedAgentsWebFetchToolConfigParams object`1059210593          Configuration override for the web_fetch tool.1059410595          - `type: optional "web_fetch"`1059610597          - `name: "web_fetch"`1059810599            Must be "web_fetch".1060010601          - `allowed_domains: optional array of string`1060210603            Only fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.1060410605          - `blocked_domains: optional array of string`1060610607            Never fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.1060810609          - `enabled: optional boolean or null`1061010611            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1061210613          - `max_content_tokens: optional number or null`1061410615            Maximum number of tokens of fetched text content to include in context per call. Does not apply to binary content such as PDFs.1061610617            format: int321061810619          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1062010621            Permission policy for tool execution.1062210623            - `BetaManagedAgentsAlwaysAllowPolicy object`1062410625              Tool calls are automatically approved without user confirmation.1062610627            - `BetaManagedAgentsAlwaysAskPolicy object`1062810629              Tool calls require user confirmation before execution.1063010631            - `BetaManagedAgentsAutoPolicy object`1063210633              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1063410635        - `BetaManagedAgentsWebSearchToolConfigParams object`1063610637          Configuration override for the web_search tool.1063810639          - `type: optional "web_search"`1064010641          - `name: "web_search"`1064210643            Must be "web_search".1064410645          - `allowed_domains: optional array of string`1064610647            Only return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.1064810649          - `blocked_domains: optional array of string`1065010651            Never return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.1065210653          - `enabled: optional boolean or null`1065410655            Whether this tool is enabled and available to Claude. Overrides the default_config setting.1065610657          - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1065810659            Permission policy for tool execution.1066010661            - `BetaManagedAgentsAlwaysAllowPolicy object`1066210663              Tool calls are automatically approved without user confirmation.1066410665            - `BetaManagedAgentsAlwaysAskPolicy object`1066610667              Tool calls require user confirmation before execution.1066810669            - `BetaManagedAgentsAutoPolicy object`1067010671              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1067210673          - `user_location: optional BetaManagedAgentsUserLocation or null`1067410675            Approximate user location for search result localization.1067610677            - `type: "approximate"`1067810679              Location precision. Only "approximate" is supported.1068010681            - `city: optional string or null`1068210683              City name.1068410685              minLength: 1, maxLength: 2551068610687            - `country: optional string or null`1068810689              Two-letter ISO 3166-1 country code, uppercase.1069010691            - `region: optional string or null`1069210693              Region or state name.1069410695              minLength: 1, maxLength: 2551069610697            - `timezone: optional string or null`1069810699              IANA timezone identifier, e.g. "America/Los_Angeles".1070010701              minLength: 1, maxLength: 2551070210703      - `default_config: optional BetaManagedAgentsAgentToolsetDefaultConfigParams or null`1070410705        Default configuration for all tools in a toolset.1070610707        - `enabled: optional boolean or null`1070810709          Whether tools are enabled and available to Claude by default. Defaults to true if not specified.1071010711        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1071210713          Permission policy for tool execution.1071410715          - `BetaManagedAgentsAlwaysAllowPolicy object`1071610717            Tool calls are automatically approved without user confirmation.1071810719          - `BetaManagedAgentsAlwaysAskPolicy object`1072010721            Tool calls require user confirmation before execution.1072210723          - `BetaManagedAgentsAutoPolicy object`1072410725            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1072610727    - `BetaManagedAgentsMCPToolsetParams object`1072810729      Configuration for tools from an MCP server defined in `mcp_servers`.1073010731      - `type: "mcp_toolset"`1073210733      - `mcp_server_name: string`1073410735        Name of the MCP server. Must match a server name from the mcp_servers array. 1-255 characters.1073610737        minLength: 1, maxLength: 2551073810739      - `configs: optional array of BetaManagedAgentsMCPToolConfigParams`1074010741        Per-tool configuration overrides.1074210743        - `name: string`1074410745          Name of the MCP tool to configure. 1-128 characters.1074610747          minLength: 1, maxLength: 1281074810749        - `enabled: optional boolean or null`1075010751          Whether this tool is enabled. Overrides the `default_config` setting.1075210753        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1075410755          Permission policy for tool execution.1075610757          - `BetaManagedAgentsAlwaysAllowPolicy object`1075810759            Tool calls are automatically approved without user confirmation.1076010761          - `BetaManagedAgentsAlwaysAskPolicy object`1076210763            Tool calls require user confirmation before execution.1076410765          - `BetaManagedAgentsAutoPolicy object`1076610767            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1076810769      - `default_config: optional BetaManagedAgentsMCPToolsetDefaultConfigParams or null`1077010771        Default configuration for all tools from an MCP server.1077210773        - `enabled: optional boolean or null`1077410775          Whether tools are enabled by default. Defaults to true if not specified.1077610777        - `permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null`1077810779          Permission policy for tool execution.1078010781          - `BetaManagedAgentsAlwaysAllowPolicy object`1078210783            Tool calls are automatically approved without user confirmation.1078410785          - `BetaManagedAgentsAlwaysAskPolicy object`1078610787            Tool calls require user confirmation before execution.1078810789          - `BetaManagedAgentsAutoPolicy object`1079010791            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1079210793    - `BetaManagedAgentsCustomToolParams object`1079410795      A custom tool that is executed by the API client rather than the agent. When the agent calls this tool, an `agent.custom_tool_use` event is emitted and the session goes idle, waiting for the client to provide the result via a `user.custom_tool_result` event.1079610797      - `type: "custom"`1079810799      - `description: string`1080010801        Description of what the tool does, shown to the agent to help it decide when to use the tool.1080210803        minLength: 11080410805      - `input_schema: BetaManagedAgentsCustomToolInputSchema`1080610807        JSON Schema for custom tool input parameters.1080810809        - `type: "object"`1081010811        - `properties: optional map[unknown] or null`1081210813        - `required: optional array of string or null`1081410815      - `name: string`1081610817        Unique name for the tool. 1-128 characters; letters, digits, underscores, and hyphens.1081810819        minLength: 1, maxLength: 1281082010821### Beta Managed Agents Session Multiagent Coordinator1082210823- `BetaManagedAgentsSessionMultiagentCoordinator object`1082410825  Resolved coordinator topology with full agent definitions for each roster member.1082610827  - `type: "coordinator"`1082810829  - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`1083010831    Full `agent` definitions the coordinator may spawn as session threads.1083210833    - `BetaManagedAgentsSessionThreadAgent object`1083410835      Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.1083610837      - `type: "agent"`1083810839      - `id: string`1084010841      - `description: string or null`1084210843      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1084410845        - `type: "url"`1084610847        - `name: string`1084810849        - `url: string`1085010851      - `model: BetaManagedAgentsModelConfig`1085210853        Model identifier and configuration.1085410855        - `id: BetaManagedAgentsModel`1085610857          The model that will power your agent.1085810859          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1086010861          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`1086210863            The model that will power your agent.1086410865            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1086610867            - `"claude-fable-5-1"`1086810869              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows1087010871            - `"claude-sonnet-5"`1087210873              High-performance model for coding and agents1087410875            - `"claude-fable-5"`1087610877              Next generation of intelligence for the hardest knowledge work and coding problems1087810879            - `"claude-opus-5"`1088010881              Powerful intelligence for long-running agents and coding1088210883            - `"claude-opus-4-8"`1088410885              Powerful intelligence for long-running agents and coding1088610887            - `"claude-opus-4-7"`1088810889              Powerful intelligence for long-running agents and coding1089010891            - `"claude-opus-4-6"`1089210893              Powerful intelligence for long-running agents and coding1089410895            - `"claude-sonnet-4-6"`1089610897              Best combination of speed and intelligence1089810899            - `"claude-haiku-4-5"`1090010901              Fastest model with near-frontier intelligence1090210903            - `"claude-haiku-4-5-20251001"`1090410905              Fastest model with near-frontier intelligence1090610907            - `"claude-opus-4-5"`1090810909              Powerful intelligence for long-running agents and coding1091010911            - `"claude-opus-4-5-20251101"`1091210913              Powerful intelligence for long-running agents and coding1091410915            - `"claude-sonnet-4-5"`1091610917              High-performance model for agents and coding1091810919            - `"claude-sonnet-4-5-20250929"`1092010921              High-performance model for agents and coding1092210923          - `string`1092410925        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`1092610927          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.1092810929          - `BetaManagedAgentsEffortLow object`1093010931            Low effort. Favors latency over reasoning depth.1093210933            - `type: "low"`1093410935          - `BetaManagedAgentsEffortMedium object`1093610937            Medium effort. Balances latency and reasoning depth.1093810939            - `type: "medium"`1094010941          - `BetaManagedAgentsEffortHigh object`1094210943            High effort. Favors reasoning depth.1094410945            - `type: "high"`1094610947          - `BetaManagedAgentsEffortXhigh object`1094810949            Extra-high effort. Not all models accept this level.1095010951            - `type: "xhigh"`1095210953          - `BetaManagedAgentsEffortMax object`1095410955            Maximum effort. Favors reasoning depth over latency.1095610957            - `type: "max"`1095810959        - `inference_geo: optional string`1096010961          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.1096210963        - `speed: optional "standard" or "fast"`1096410965          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1096610967          - `"standard"`1096810969          - `"fast"`1097010971      - `name: string`1097210973      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1097410975        - `BetaManagedAgentsAnthropicSkill object`1097610977          A resolved Anthropic-managed skill.1097810979          - `type: "anthropic"`1098010981          - `skill_id: string`1098210983          - `version: string`1098410985        - `BetaManagedAgentsCustomSkill object`1098610987          A resolved user-created custom skill.1098810989          - `type: "custom"`1099010991          - `skill_id: string`1099210993          - `version: string`1099410995      - `system: string or null`1099610997      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1099810999        - `BetaManagedAgentsAgentToolset20260401 object`1100011001          - `type: "agent_toolset_20260401"`1100211003          - `configs: array of BetaManagedAgentsAgentToolConfig`1100411005            - `BetaManagedAgentsBashToolConfig object`1100611007              Configuration for the bash tool.1100811009              - `type: "bash"`1101011011              - `enabled: boolean`1101211013              - `name: "bash"`1101411015              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1101611017                Permission policy for tool execution.1101811019                - `BetaManagedAgentsAlwaysAllowPolicy object`1102011021                  Tool calls are automatically approved without user confirmation.1102211023                  - `type: "always_allow"`1102411025                - `BetaManagedAgentsAlwaysAskPolicy object`1102611027                  Tool calls require user confirmation before execution.1102811029                  - `type: "always_ask"`1103011031                - `BetaManagedAgentsAutoPolicy object`1103211033                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1103411035                  - `type: "auto"`1103611037            - `BetaManagedAgentsEditToolConfig object`1103811039              Configuration for the edit tool.1104011041              - `type: "edit"`1104211043              - `enabled: boolean`1104411045              - `name: "edit"`1104611047              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1104811049                Permission policy for tool execution.1105011051                - `BetaManagedAgentsAlwaysAllowPolicy object`1105211053                  Tool calls are automatically approved without user confirmation.1105411055                - `BetaManagedAgentsAlwaysAskPolicy object`1105611057                  Tool calls require user confirmation before execution.1105811059                - `BetaManagedAgentsAutoPolicy object`1106011061                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1106211063            - `BetaManagedAgentsReadToolConfig object`1106411065              Configuration for the read tool.1106611067              - `type: "read"`1106811069              - `enabled: boolean`1107011071              - `name: "read"`1107211073              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1107411075                Permission policy for tool execution.1107611077                - `BetaManagedAgentsAlwaysAllowPolicy object`1107811079                  Tool calls are automatically approved without user confirmation.1108011081                - `BetaManagedAgentsAlwaysAskPolicy object`1108211083                  Tool calls require user confirmation before execution.1108411085                - `BetaManagedAgentsAutoPolicy object`1108611087                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1108811089            - `BetaManagedAgentsWriteToolConfig object`1109011091              Configuration for the write tool.1109211093              - `type: "write"`1109411095              - `enabled: boolean`1109611097              - `name: "write"`1109811099              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1110011101                Permission policy for tool execution.1110211103                - `BetaManagedAgentsAlwaysAllowPolicy object`1110411105                  Tool calls are automatically approved without user confirmation.1110611107                - `BetaManagedAgentsAlwaysAskPolicy object`1110811109                  Tool calls require user confirmation before execution.1111011111                - `BetaManagedAgentsAutoPolicy object`1111211113                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1111411115            - `BetaManagedAgentsGlobToolConfig object`1111611117              Configuration for the glob tool.1111811119              - `type: "glob"`1112011121              - `enabled: boolean`1112211123              - `name: "glob"`1112411125              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1112611127                Permission policy for tool execution.1112811129                - `BetaManagedAgentsAlwaysAllowPolicy object`1113011131                  Tool calls are automatically approved without user confirmation.1113211133                - `BetaManagedAgentsAlwaysAskPolicy object`1113411135                  Tool calls require user confirmation before execution.1113611137                - `BetaManagedAgentsAutoPolicy object`1113811139                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1114011141            - `BetaManagedAgentsGrepToolConfig object`1114211143              Configuration for the grep tool.1114411145              - `type: "grep"`1114611147              - `enabled: boolean`1114811149              - `name: "grep"`1115011151              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1115211153                Permission policy for tool execution.1115411155                - `BetaManagedAgentsAlwaysAllowPolicy object`1115611157                  Tool calls are automatically approved without user confirmation.1115811159                - `BetaManagedAgentsAlwaysAskPolicy object`1116011161                  Tool calls require user confirmation before execution.1116211163                - `BetaManagedAgentsAutoPolicy object`1116411165                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1116611167            - `BetaManagedAgentsWebFetchToolConfig object`1116811169              Configuration for the web_fetch tool.1117011171              - `type: "web_fetch"`1117211173              - `enabled: boolean`1117411175              - `name: "web_fetch"`1117611177              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1117811179                Permission policy for tool execution.1118011181                - `BetaManagedAgentsAlwaysAllowPolicy object`1118211183                  Tool calls are automatically approved without user confirmation.1118411185                - `BetaManagedAgentsAlwaysAskPolicy object`1118611187                  Tool calls require user confirmation before execution.1118811189                - `BetaManagedAgentsAutoPolicy object`1119011191                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1119211193              - `allowed_domains: optional array of string`1119411195              - `blocked_domains: optional array of string`1119611197              - `max_content_tokens: optional number or null`1119811199                format: int321120011201            - `BetaManagedAgentsWebSearchToolConfig object`1120211203              Configuration for the web_search tool.1120411205              - `type: "web_search"`1120611207              - `enabled: boolean`1120811209              - `name: "web_search"`1121011211              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1121211213                Permission policy for tool execution.1121411215                - `BetaManagedAgentsAlwaysAllowPolicy object`1121611217                  Tool calls are automatically approved without user confirmation.1121811219                - `BetaManagedAgentsAlwaysAskPolicy object`1122011221                  Tool calls require user confirmation before execution.1122211223                - `BetaManagedAgentsAutoPolicy object`1122411225                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1122611227              - `allowed_domains: optional array of string`1122811229              - `blocked_domains: optional array of string`1123011231              - `user_location: optional BetaManagedAgentsUserLocation or null`1123211233                Approximate user location for search result localization.1123411235                - `type: "approximate"`1123611237                  Location precision. Only "approximate" is supported.1123811239                - `city: optional string or null`1124011241                  City name.1124211243                  minLength: 1, maxLength: 2551124411245                - `country: optional string or null`1124611247                  Two-letter ISO 3166-1 country code, uppercase.1124811249                - `region: optional string or null`1125011251                  Region or state name.1125211253                  minLength: 1, maxLength: 2551125411255                - `timezone: optional string or null`1125611257                  IANA timezone identifier, e.g. "America/Los_Angeles".1125811259                  minLength: 1, maxLength: 2551126011261          - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`1126211263            Resolved default configuration for agent tools.1126411265            - `enabled: boolean`1126611267            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1126811269              Permission policy for tool execution.1127011271              - `BetaManagedAgentsAlwaysAllowPolicy object`1127211273                Tool calls are automatically approved without user confirmation.1127411275              - `BetaManagedAgentsAlwaysAskPolicy object`1127611277                Tool calls require user confirmation before execution.1127811279              - `BetaManagedAgentsAutoPolicy object`1128011281                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1128211283        - `BetaManagedAgentsMCPToolset object`1128411285          - `type: "mcp_toolset"`1128611287          - `configs: array of BetaManagedAgentsMCPToolConfig`1128811289            - `enabled: boolean`1129011291            - `name: string`1129211293            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1129411295              Permission policy for tool execution.1129611297              - `BetaManagedAgentsAlwaysAllowPolicy object`1129811299                Tool calls are automatically approved without user confirmation.1130011301              - `BetaManagedAgentsAlwaysAskPolicy object`1130211303                Tool calls require user confirmation before execution.1130411305              - `BetaManagedAgentsAutoPolicy object`1130611307                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1130811309          - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`1131011311            Resolved default configuration for all tools from an MCP server.1131211313            - `enabled: boolean`1131411315            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1131611317              Permission policy for tool execution.1131811319              - `BetaManagedAgentsAlwaysAllowPolicy object`1132011321                Tool calls are automatically approved without user confirmation.1132211323              - `BetaManagedAgentsAlwaysAskPolicy object`1132411325                Tool calls require user confirmation before execution.1132611327              - `BetaManagedAgentsAutoPolicy object`1132811329                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1133011331          - `mcp_server_name: string`1133211333        - `BetaManagedAgentsCustomTool object`1133411335          A custom tool as returned in API responses.1133611337          - `type: "custom"`1133811339          - `description: string`1134011341          - `input_schema: BetaManagedAgentsCustomToolInputSchema`1134211343            JSON Schema for custom tool input parameters.1134411345            - `type: "object"`1134611347            - `properties: optional map[unknown] or null`1134811349            - `required: optional array of string or null`1135011351          - `name: string`1135211353      - `version: number`1135411355        format: int321135611357    - `BetaManagedAgentsAdvisor object`1135811359      Platform advisor roster entry: a model the session's primary thread may consult mid-turn.1136011361      - `type: "advisor"`1136211363      - `model: string`1136411365        The advisor model id.1136611367### Beta Managed Agents Session Stats1136811369- `BetaManagedAgentsSessionStats object`1137011371  Timing statistics for a session.1137211373  - `active_seconds: optional number`1137411375    Cumulative time in seconds the session spent in `running` status. Excludes idle time.1137611377    format: double1137811379  - `duration_seconds: optional number`1138011381    Elapsed time since session creation in seconds. For terminated sessions, frozen at the final update.1138211383    format: double1138411385### Beta Managed Agents Session Updated Event1138611387- `BetaManagedAgentsSessionUpdatedEvent object`1138811389  Emitted when an UpdateSession request changed at least one field. Carries only the fields that changed; absent fields were not part of the update. The new configuration applies from the next turn.1139011391  - `type: "session.updated"`1139211393  - `id: string`1139411395    Unique identifier for this event.1139611397  - `processed_at: string`1139811399    A timestamp in RFC 3339 format1140011401    format: date-time1140211403  - `agent: optional BetaManagedAgentsSessionAgent or null`1140411405    Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.1140611407    - `type: "agent"`1140811409    - `id: string`1141011411    - `description: string or null`1141211413    - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1141411415      - `type: "url"`1141611417      - `name: string`1141811419      - `url: string`1142011421    - `model: BetaManagedAgentsModelConfig`1142211423      Model identifier and configuration.1142411425      - `id: BetaManagedAgentsModel`1142611427        The model that will power your agent.1142811429        See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1143011431        - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`1143211433          The model that will power your agent.1143411435          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1143611437          - `"claude-fable-5-1"`1143811439            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows1144011441          - `"claude-sonnet-5"`1144211443            High-performance model for coding and agents1144411445          - `"claude-fable-5"`1144611447            Next generation of intelligence for the hardest knowledge work and coding problems1144811449          - `"claude-opus-5"`1145011451            Powerful intelligence for long-running agents and coding1145211453          - `"claude-opus-4-8"`1145411455            Powerful intelligence for long-running agents and coding1145611457          - `"claude-opus-4-7"`1145811459            Powerful intelligence for long-running agents and coding1146011461          - `"claude-opus-4-6"`1146211463            Powerful intelligence for long-running agents and coding1146411465          - `"claude-sonnet-4-6"`1146611467            Best combination of speed and intelligence1146811469          - `"claude-haiku-4-5"`1147011471            Fastest model with near-frontier intelligence1147211473          - `"claude-haiku-4-5-20251001"`1147411475            Fastest model with near-frontier intelligence1147611477          - `"claude-opus-4-5"`1147811479            Powerful intelligence for long-running agents and coding1148011481          - `"claude-opus-4-5-20251101"`1148211483            Powerful intelligence for long-running agents and coding1148411485          - `"claude-sonnet-4-5"`1148611487            High-performance model for agents and coding1148811489          - `"claude-sonnet-4-5-20250929"`1149011491            High-performance model for agents and coding1149211493        - `string`1149411495      - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`1149611497        How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.1149811499        - `BetaManagedAgentsEffortLow object`1150011501          Low effort. Favors latency over reasoning depth.1150211503          - `type: "low"`1150411505        - `BetaManagedAgentsEffortMedium object`1150611507          Medium effort. Balances latency and reasoning depth.1150811509          - `type: "medium"`1151011511        - `BetaManagedAgentsEffortHigh object`1151211513          High effort. Favors reasoning depth.1151411515          - `type: "high"`1151611517        - `BetaManagedAgentsEffortXhigh object`1151811519          Extra-high effort. Not all models accept this level.1152011521          - `type: "xhigh"`1152211523        - `BetaManagedAgentsEffortMax object`1152411525          Maximum effort. Favors reasoning depth over latency.1152611527          - `type: "max"`1152811529      - `inference_geo: optional string`1153011531        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.1153211533      - `speed: optional "standard" or "fast"`1153411535        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1153611537        - `"standard"`1153811539        - `"fast"`1154011541    - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`1154211543      Resolved coordinator topology with full agent definitions for each roster member.1154411545      - `type: "coordinator"`1154611547      - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`1154811549        Full `agent` definitions the coordinator may spawn as session threads.1155011551        - `BetaManagedAgentsSessionThreadAgent object`1155211553          Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.1155411555          - `type: "agent"`1155611557          - `id: string`1155811559          - `description: string or null`1156011561          - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1156211563            - `type: "url"`1156411565            - `name: string`1156611567            - `url: string`1156811569          - `model: BetaManagedAgentsModelConfig`1157011571            Model identifier and configuration.1157211573          - `name: string`1157411575          - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1157611577            - `BetaManagedAgentsAnthropicSkill object`1157811579              A resolved Anthropic-managed skill.1158011581              - `type: "anthropic"`1158211583              - `skill_id: string`1158411585              - `version: string`1158611587            - `BetaManagedAgentsCustomSkill object`1158811589              A resolved user-created custom skill.1159011591              - `type: "custom"`1159211593              - `skill_id: string`1159411595              - `version: string`1159611597          - `system: string or null`1159811599          - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1160011601            - `BetaManagedAgentsAgentToolset20260401 object`1160211603              - `type: "agent_toolset_20260401"`1160411605              - `configs: array of BetaManagedAgentsAgentToolConfig`1160611607                - `BetaManagedAgentsBashToolConfig object`1160811609                  Configuration for the bash tool.1161011611                  - `type: "bash"`1161211613                  - `enabled: boolean`1161411615                  - `name: "bash"`1161611617                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1161811619                    Permission policy for tool execution.1162011621                    - `BetaManagedAgentsAlwaysAllowPolicy object`1162211623                      Tool calls are automatically approved without user confirmation.1162411625                      - `type: "always_allow"`1162611627                    - `BetaManagedAgentsAlwaysAskPolicy object`1162811629                      Tool calls require user confirmation before execution.1163011631                      - `type: "always_ask"`1163211633                    - `BetaManagedAgentsAutoPolicy object`1163411635                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1163611637                      - `type: "auto"`1163811639                - `BetaManagedAgentsEditToolConfig object`1164011641                  Configuration for the edit tool.1164211643                  - `type: "edit"`1164411645                  - `enabled: boolean`1164611647                  - `name: "edit"`1164811649                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1165011651                    Permission policy for tool execution.1165211653                    - `BetaManagedAgentsAlwaysAllowPolicy object`1165411655                      Tool calls are automatically approved without user confirmation.1165611657                    - `BetaManagedAgentsAlwaysAskPolicy object`1165811659                      Tool calls require user confirmation before execution.1166011661                    - `BetaManagedAgentsAutoPolicy object`1166211663                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1166411665                - `BetaManagedAgentsReadToolConfig object`1166611667                  Configuration for the read tool.1166811669                  - `type: "read"`1167011671                  - `enabled: boolean`1167211673                  - `name: "read"`1167411675                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1167611677                    Permission policy for tool execution.1167811679                    - `BetaManagedAgentsAlwaysAllowPolicy object`1168011681                      Tool calls are automatically approved without user confirmation.1168211683                    - `BetaManagedAgentsAlwaysAskPolicy object`1168411685                      Tool calls require user confirmation before execution.1168611687                    - `BetaManagedAgentsAutoPolicy object`1168811689                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1169011691                - `BetaManagedAgentsWriteToolConfig object`1169211693                  Configuration for the write tool.1169411695                  - `type: "write"`1169611697                  - `enabled: boolean`1169811699                  - `name: "write"`1170011701                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1170211703                    Permission policy for tool execution.1170411705                    - `BetaManagedAgentsAlwaysAllowPolicy object`1170611707                      Tool calls are automatically approved without user confirmation.1170811709                    - `BetaManagedAgentsAlwaysAskPolicy object`1171011711                      Tool calls require user confirmation before execution.1171211713                    - `BetaManagedAgentsAutoPolicy object`1171411715                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1171611717                - `BetaManagedAgentsGlobToolConfig object`1171811719                  Configuration for the glob tool.1172011721                  - `type: "glob"`1172211723                  - `enabled: boolean`1172411725                  - `name: "glob"`1172611727                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1172811729                    Permission policy for tool execution.1173011731                    - `BetaManagedAgentsAlwaysAllowPolicy object`1173211733                      Tool calls are automatically approved without user confirmation.1173411735                    - `BetaManagedAgentsAlwaysAskPolicy object`1173611737                      Tool calls require user confirmation before execution.1173811739                    - `BetaManagedAgentsAutoPolicy object`1174011741                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1174211743                - `BetaManagedAgentsGrepToolConfig object`1174411745                  Configuration for the grep tool.1174611747                  - `type: "grep"`1174811749                  - `enabled: boolean`1175011751                  - `name: "grep"`1175211753                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1175411755                    Permission policy for tool execution.1175611757                    - `BetaManagedAgentsAlwaysAllowPolicy object`1175811759                      Tool calls are automatically approved without user confirmation.1176011761                    - `BetaManagedAgentsAlwaysAskPolicy object`1176211763                      Tool calls require user confirmation before execution.1176411765                    - `BetaManagedAgentsAutoPolicy object`1176611767                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1176811769                - `BetaManagedAgentsWebFetchToolConfig object`1177011771                  Configuration for the web_fetch tool.1177211773                  - `type: "web_fetch"`1177411775                  - `enabled: boolean`1177611777                  - `name: "web_fetch"`1177811779                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1178011781                    Permission policy for tool execution.1178211783                    - `BetaManagedAgentsAlwaysAllowPolicy object`1178411785                      Tool calls are automatically approved without user confirmation.1178611787                    - `BetaManagedAgentsAlwaysAskPolicy object`1178811789                      Tool calls require user confirmation before execution.1179011791                    - `BetaManagedAgentsAutoPolicy object`1179211793                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1179411795                  - `allowed_domains: optional array of string`1179611797                  - `blocked_domains: optional array of string`1179811799                  - `max_content_tokens: optional number or null`1180011801                    format: int321180211803                - `BetaManagedAgentsWebSearchToolConfig object`1180411805                  Configuration for the web_search tool.1180611807                  - `type: "web_search"`1180811809                  - `enabled: boolean`1181011811                  - `name: "web_search"`1181211813                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1181411815                    Permission policy for tool execution.1181611817                    - `BetaManagedAgentsAlwaysAllowPolicy object`1181811819                      Tool calls are automatically approved without user confirmation.1182011821                    - `BetaManagedAgentsAlwaysAskPolicy object`1182211823                      Tool calls require user confirmation before execution.1182411825                    - `BetaManagedAgentsAutoPolicy object`1182611827                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1182811829                  - `allowed_domains: optional array of string`1183011831                  - `blocked_domains: optional array of string`1183211833                  - `user_location: optional BetaManagedAgentsUserLocation or null`1183411835                    Approximate user location for search result localization.1183611837                    - `type: "approximate"`1183811839                      Location precision. Only "approximate" is supported.1184011841                    - `city: optional string or null`1184211843                      City name.1184411845                      minLength: 1, maxLength: 2551184611847                    - `country: optional string or null`1184811849                      Two-letter ISO 3166-1 country code, uppercase.1185011851                    - `region: optional string or null`1185211853                      Region or state name.1185411855                      minLength: 1, maxLength: 2551185611857                    - `timezone: optional string or null`1185811859                      IANA timezone identifier, e.g. "America/Los_Angeles".1186011861                      minLength: 1, maxLength: 2551186211863              - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`1186411865                Resolved default configuration for agent tools.1186611867                - `enabled: boolean`1186811869                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1187011871                  Permission policy for tool execution.1187211873                  - `BetaManagedAgentsAlwaysAllowPolicy object`1187411875                    Tool calls are automatically approved without user confirmation.1187611877                  - `BetaManagedAgentsAlwaysAskPolicy object`1187811879                    Tool calls require user confirmation before execution.1188011881                  - `BetaManagedAgentsAutoPolicy object`1188211883                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1188411885            - `BetaManagedAgentsMCPToolset object`1188611887              - `type: "mcp_toolset"`1188811889              - `configs: array of BetaManagedAgentsMCPToolConfig`1189011891                - `enabled: boolean`1189211893                - `name: string`1189411895                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1189611897                  Permission policy for tool execution.1189811899                  - `BetaManagedAgentsAlwaysAllowPolicy object`1190011901                    Tool calls are automatically approved without user confirmation.1190211903                  - `BetaManagedAgentsAlwaysAskPolicy object`1190411905                    Tool calls require user confirmation before execution.1190611907                  - `BetaManagedAgentsAutoPolicy object`1190811909                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1191011911              - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`1191211913                Resolved default configuration for all tools from an MCP server.1191411915                - `enabled: boolean`1191611917                - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1191811919                  Permission policy for tool execution.1192011921                  - `BetaManagedAgentsAlwaysAllowPolicy object`1192211923                    Tool calls are automatically approved without user confirmation.1192411925                  - `BetaManagedAgentsAlwaysAskPolicy object`1192611927                    Tool calls require user confirmation before execution.1192811929                  - `BetaManagedAgentsAutoPolicy object`1193011931                    The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1193211933              - `mcp_server_name: string`1193411935            - `BetaManagedAgentsCustomTool object`1193611937              A custom tool as returned in API responses.1193811939              - `type: "custom"`1194011941              - `description: string`1194211943              - `input_schema: BetaManagedAgentsCustomToolInputSchema`1194411945                JSON Schema for custom tool input parameters.1194611947                - `type: "object"`1194811949                - `properties: optional map[unknown] or null`1195011951                - `required: optional array of string or null`1195211953              - `name: string`1195411955          - `version: number`1195611957            format: int321195811959        - `BetaManagedAgentsAdvisor object`1196011961          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.1196211963          - `type: "advisor"`1196411965          - `model: string`1196611967            The advisor model id.1196811969    - `name: string`1197011971    - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1197211973      - `BetaManagedAgentsAnthropicSkill object`1197411975        A resolved Anthropic-managed skill.1197611977      - `BetaManagedAgentsCustomSkill object`1197811979        A resolved user-created custom skill.1198011981    - `system: string or null`1198211983    - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1198411985      - `BetaManagedAgentsAgentToolset20260401 object`1198611987      - `BetaManagedAgentsMCPToolset object`1198811989      - `BetaManagedAgentsCustomTool object`1199011991        A custom tool as returned in API responses.1199211993    - `version: number`1199411995      format: int321199611997  - `budget: optional BetaManagedAgentsBudgetLimit or null`1199811999    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1200012001    - `type: "limit"`1200212003    - `max_list_cost: BetaMonetaryAmount`1200412005      A monetary amount in a specific currency.1200612007      - `amount: string`1200812009        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.1201012011      - `currency: BetaCurrency`1201212013        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.1201412015  - `metadata: optional map[string]`1201612017    The session's full metadata bag after the update. Present when the update set non-empty metadata; absent when metadata was unchanged or cleared to empty.1201812019  - `title: optional string or null`1202012021    The session's new title. Present only when the update changed it.1202212023### Beta Managed Agents Session Usage1202412025- `BetaManagedAgentsSessionUsage object`1202612027  Cumulative token usage for a session across all turns.1202812029  - `active_seconds: optional number`1203012031    Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once, unlike `stats.active_seconds`, which sums each thread's own active time. This is the duration the session's runtime cost is priced on.1203212033    format: double1203412035  - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`1203612037    Prompt-cache creation token usage broken down by cache lifetime.1203812039    - `ephemeral_1h_input_tokens: optional number`1204012041      Tokens used to create 1-hour ephemeral cache entries.1204212043      format: int321204412045    - `ephemeral_5m_input_tokens: optional number`1204612047      Tokens used to create 5-minute ephemeral cache entries.1204812049      format: int321205012051  - `cache_read_input_tokens: optional number`1205212053    Total tokens read from prompt cache.1205412055    format: int321205612057  - `input_tokens: optional number`1205812059    Total input tokens consumed across all turns.1206012061    format: int321206212063  - `list_cost: optional BetaMonetaryAmount or null`1206412065    A monetary amount in a specific currency.1206612067    - `amount: string`1206812069      Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.1207012071    - `currency: BetaCurrency`1207212073      Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.1207412075  - `output_tokens: optional number`1207612077    Total output tokens generated across all turns.1207812079    format: int321208012081  - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`1208212083    Cumulative count of server-executed tool invocations, broken down by tool.1208412085    - `web_fetch_requests: optional number`1208612087      Number of server-executed web fetch requests.1208812089      format: int321209012091    - `web_search_requests: optional number`1209212093      Number of server-executed web search requests.1209412095      format: int321209612097### Beta Managed Agents Session Usage Event1209812099- `BetaManagedAgentsSessionUsageEvent object`1210012101  Periodic snapshot of the session's cumulative usage and tracked list cost.1210212103  - `type: "session.usage"`1210412105  - `id: string`1210612107    Unique identifier for this event.1210812109  - `processed_at: string`1211012111    A timestamp in RFC 3339 format1211212113    format: date-time1211412115  - `usage: BetaManagedAgentsSessionUsageSnapshot`1211612117    Point-in-time snapshot of a session's cumulative usage.1211812119    - `active_seconds: optional number`1212012121      Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once. This is the duration the session's runtime cost is priced on.1212212123      format: double1212412125    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`1212612127      Prompt-cache creation token usage broken down by cache lifetime.1212812129      - `ephemeral_1h_input_tokens: optional number`1213012131        Tokens used to create 1-hour ephemeral cache entries.1213212133        format: int321213412135      - `ephemeral_5m_input_tokens: optional number`1213612137        Tokens used to create 5-minute ephemeral cache entries.1213812139        format: int321214012141    - `cache_read_input_tokens: optional number`1214212143      Total tokens read from prompt cache.1214412145      format: int321214612147    - `input_tokens: optional number`1214812149      Total input tokens consumed across all turns.1215012151      format: int321215212153    - `list_cost: optional BetaMonetaryAmount`1215412155      A monetary amount in a specific currency.1215612157      - `amount: string`1215812159        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.1216012161      - `currency: BetaCurrency`1216212163        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.1216412165    - `output_tokens: optional number`1216612167      Total output tokens generated across all turns.1216812169      format: int321217012171    - `server_tool_use: optional BetaManagedAgentsServerToolUsage`1217212173      Cumulative count of server-executed tool invocations, broken down by tool.1217412175      - `web_fetch_requests: optional number`1217612177        Number of server-executed web fetch requests.1217812179        format: int321218012181      - `web_search_requests: optional number`1218212183        Number of server-executed web search requests.1218412185        format: int321218612187  - `budget: optional BetaManagedAgentsBudgetLimit or null`1218812189    A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1219012191    - `type: "limit"`1219212193    - `max_list_cost: BetaMonetaryAmount`1219412195      A monetary amount in a specific currency.1219612197### Beta Managed Agents Start Event1219812199- `BetaManagedAgentsStartEvent object`1220012201  Opens a preview of a buffered event. Carries the previewed event's type and id only. Followed by zero or more event_delta events with the same event id, normally concluded by the buffered event carrying that id. If the producing model request ends without that event (an error or interrupt mid-stream), its terminal span.model_request_end closes the preview. Only sent on stream connections that opt in via event_deltas; never appears in event history.1220212203  - `type: "event_start"`1220412205  - `event: BetaManagedAgentsStartEventPreview`1220612207    The previewed event's type and id. The event type determines which delta types the preview's event_delta events carry: agent.message events stream content_delta fragments; agent.thinking previews are start-only — no deltas follow, and the buffered agent.thinking with the same id concludes them.1220812209    - `BetaManagedAgentsAgentMessagePreview object`1221012211      - `type: "agent.message"`1221212213      - `id: string`1221412215        The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.1221612217    - `BetaManagedAgentsAgentThinkingPreview object`1221812219      - `type: "agent.thinking"`1222012221      - `id: string`1222212223        The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.1222412225### Beta Managed Agents Start Event Preview1222612227- `BetaManagedAgentsStartEventPreview = BetaManagedAgentsAgentMessagePreview or BetaManagedAgentsAgentThinkingPreview`1222812229  - `BetaManagedAgentsAgentMessagePreview object`1223012231    - `type: "agent.message"`1223212233    - `id: string`1223412235      The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.1223612237  - `BetaManagedAgentsAgentThinkingPreview object`1223812239    - `type: "agent.thinking"`1224012241    - `id: string`1224212243      The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.1224412245### Beta Managed Agents System Content Block1224612247- `BetaManagedAgentsSystemContentBlock object`1224812249  Regular text content.1225012251  - `type: "text"`1225212253  - `text: string`1225412255    The text content.1225612257    minLength: 11225812259### Beta Managed Agents System Message Event1226012261- `BetaManagedAgentsSystemMessageEvent object`1226212263  A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.1226412265  - `type: "system.message"`1226612267  - `id: string`1226812269    Unique identifier for this event.1227012271  - `content: array of BetaManagedAgentsSystemContentBlock`1227212273    System content blocks. Text-only.1227412275    - `type: "text"`1227612277    - `text: string`1227812279      The text content.1228012281      minLength: 11228212283  - `processed_at: optional string or null`1228412285    A timestamp in RFC 3339 format1228612287    format: date-time1228812289### Beta Managed Agents User Tool Result Event1229012291- `BetaManagedAgentsUserToolResultEvent object`1229212293  Event sent by the client providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.1229412295  - `type: "user.tool_result"`1229612297  - `id: string`1229812299    Unique identifier for this event.1230012301  - `tool_use_id: string`1230212303    The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1230412305  - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1230612307    The result content returned by the tool.1230812309    - `BetaManagedAgentsTextBlock object`1231012311      Regular text content.1231212313      - `type: "text"`1231412315      - `text: string`1231612317        The text content.1231812319        minLength: 11232012321    - `BetaManagedAgentsImageBlock object`1232212323      Image content specified directly as base64 data or as a reference via a URL.1232412325      - `type: "image"`1232612327      - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`1232812329        Union type for image source variants.1233012331        - `BetaManagedAgentsBase64ImageSource object`1233212333          Base64-encoded image data.1233412335          - `type: "base64"`1233612337          - `data: string`1233812339            Base64-encoded image data.1234012341            minLength: 11234212343          - `media_type: string`1234412345            MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").1234612347            minLength: 11234812349        - `BetaManagedAgentsURLImageSource object`1235012351          Image referenced by URL.1235212353          - `type: "url"`1235412355          - `url: string`1235612357            URL of the image to fetch.1235812359            minLength: 11236012361        - `BetaManagedAgentsFileImageSource object`1236212363          Image referenced by file ID.1236412365          - `type: "file"`1236612367          - `file_id: string`1236812369            ID of a previously uploaded file.1237012371            minLength: 11237212373    - `BetaManagedAgentsDocumentBlock object`1237412375      Document content, either specified directly as base64 data, as text, or as a reference via a URL.1237612377      - `type: "document"`1237812379      - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`1238012381        Union type for document source variants.1238212383        - `BetaManagedAgentsBase64DocumentSource object`1238412385          Base64-encoded document data.1238612387          - `type: "base64"`1238812389          - `data: string`1239012391            Base64-encoded document data.1239212393            minLength: 11239412395          - `media_type: string`1239612397            MIME type of the document (e.g., "application/pdf").1239812399            minLength: 11240012401        - `BetaManagedAgentsPlainTextDocumentSource object`1240212403          Plain text document content.1240412405          - `type: "text"`1240612407          - `data: string`1240812409            The plain text content.1241012411            minLength: 11241212413          - `media_type: "text/plain"`1241412415            MIME type of the text content. Must be "text/plain".1241612417        - `BetaManagedAgentsURLDocumentSource object`1241812419          Document referenced by URL.1242012421          - `type: "url"`1242212423          - `url: string`1242412425            URL of the document to fetch.1242612427            minLength: 11242812429        - `BetaManagedAgentsFileDocumentSource object`1243012431          Document referenced by file ID.1243212433          - `type: "file"`1243412435          - `file_id: string`1243612437            ID of a previously uploaded file.1243812439            minLength: 11244012441      - `context: optional string or null`1244212443        Additional context about the document for the model.1244412445      - `title: optional string or null`1244612447        The title of the document.1244812449    - `BetaManagedAgentsSearchResultBlock object`1245012451      A block containing a web search result.1245212453      - `type: "search_result"`1245412455      - `citations: BetaManagedAgentsSearchResultCitations`1245612457        Citation settings for a search result.1245812459        - `enabled: boolean`1246012461          Whether citations are enabled for this search result.1246212463      - `content: array of BetaManagedAgentsSearchResultContent`1246412465        Array of text content blocks from the search result.1246612467        - `type: "text"`1246812469        - `text: string`1247012471          The text content.1247212473          minLength: 11247412475      - `source: string`1247612477        The URL source of the search result.1247812479        minLength: 11248012481      - `title: string`1248212483        The title of the search result.1248412485        minLength: 11248612487  - `is_error: optional boolean or null`1248812489    Whether the tool execution resulted in an error.1249012491  - `processed_at: optional string or null`1249212493    A timestamp in RFC 3339 format1249412495    format: date-time1249612497  - `session_thread_id: optional string or null`1249812499    Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1250012501## Sessions › Events1250212503### List Events1250412505**GET** `/v1/sessions/{session_id}/events`1250612507List Events1250812509#### Path parameters1251012511- `session_id: string`1251212513#### Query parameters1251412515- `"created_at[gt]": optional string`1251612517  Return events created after this time (exclusive). Compared against the event's `processed_at` value.1251812519  format: date-time1252012521- `"created_at[gte]": optional string`1252212523  Return events created at or after this time (inclusive). Compared against the event's `processed_at` value.1252412525  format: date-time1252612527- `"created_at[lt]": optional string`1252812529  Return events created before this time (exclusive). Compared against the event's `processed_at` value.1253012531  format: date-time1253212533- `"created_at[lte]": optional string`1253412535  Return events created at or before this time (inclusive). Compared against the event's `processed_at` value.1253612537  format: date-time1253812539- `limit: optional number`1254012541  Query parameter for limit1254212543  format: int321254412545- `order: optional "asc" or "desc"`1254612547  Sort direction for results, ordered by the event's `processed_at`. Defaults to `asc` (chronological).1254812549  - `"asc"`1255012551  - `"desc"`1255212553- `page: optional string`1255412555  Opaque pagination cursor from a previous response's `next_page`.1255612557- `types: optional array of string`1255812559  Filter by event type. Values match the `type` field on returned events (for example, `user.message` or `agent.tool_use`). Omit to return all event types.1256012561#### Headers1256212563- `"anthropic-beta": optional array of AnthropicBeta`1256412565  Optional header to specify the beta version(s) you want to use.1256612567  - `string`1256812569  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1257012571    - `"message-batches-2024-09-24"`1257212573    - `"prompt-caching-2024-07-31"`1257412575    - `"computer-use-2024-10-22"`1257612577    - `"computer-use-2025-01-24"`1257812579    - `"pdfs-2024-09-25"`1258012581    - `"token-counting-2024-11-01"`1258212583    - `"token-efficient-tools-2025-02-19"`1258412585    - `"output-128k-2025-02-19"`1258612587    - `"files-api-2025-04-14"`1258812589    - `"mcp-client-2025-04-04"`1259012591    - `"mcp-client-2025-11-20"`1259212593    - `"dev-full-thinking-2025-05-14"`1259412595    - `"interleaved-thinking-2025-05-14"`1259612597    - `"code-execution-2025-05-22"`1259812599    - `"extended-cache-ttl-2025-04-11"`1260012601    - `"context-1m-2025-08-07"`1260212603    - `"context-management-2025-06-27"`1260412605    - `"model-context-window-exceeded-2025-08-26"`1260612607    - `"skills-2025-10-02"`1260812609    - `"fast-mode-2026-02-01"`1261012611    - `"output-300k-2026-03-24"`1261212613    - `"user-profiles-2026-03-24"`1261412615    - `"user-profiles-2026-08-18"`1261612617    - `"user-profiles-2026-09-04"`1261812619    - `"advisor-tool-2026-03-01"`1262012621    - `"managed-agents-2026-04-01"`1262212623    - `"cache-diagnosis-2026-04-07"`1262412625    - `"dreaming-2026-04-21"`1262612627    - `"thinking-token-count-2026-05-13"`1262812629    - `"server-side-fallback-2026-06-01"`1263012631    - `"server-side-fallback-2026-07-01"`1263212633    - `"fallback-credit-2026-06-01"`1263412635    - `"fallback-credit-2026-07-01"`1263612637    - `"agent-memory-2026-07-22"`1263812639    - `"mid-conversation-tool-changes-2026-07-01"`1264012641    - `"compact-2026-01-12"`1264212643    - `"computer-use-2025-11-24"`1264412645    - `"mcp-tunnels-2026-06-22"`1264612647    - `"structured-outputs-2025-11-13"`1264812649    - `"task-budgets-2026-03-13"`1265012651    - `"thinking-display-updates-2026-08-18"`1265212653    - `"ce-user-management-2026-07-13"`1265412655    - `"mid-conversation-output-config-2026-07-01"`1265612657    - `"thinking-binding-controls-2026-08-01"`1265812659    - `"mid-conversation-system-clear-at-2026-08-21"`1266012661    - `"compact-2026-09-04"`1266212663- `"anthropic-workspace-id": optional string`1266412665#### Returns1266612667- `data: optional array of BetaManagedAgentsSessionEvent`1266812669  Events for the session, ordered by `processed_at`.1267012671  - `BetaManagedAgentsUserMessageEvent object`1267212673    A user message event in the session conversation.1267412675    - `type: "user.message"`1267612677    - `id: string`1267812679      Unique identifier for this event.1268012681    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1268212683      Array of content blocks comprising the user message.1268412685      - `BetaManagedAgentsTextBlock object`1268612687        Regular text content.1268812689        - `type: "text"`1269012691        - `text: string`1269212693          The text content.1269412695          minLength: 11269612697      - `BetaManagedAgentsImageBlock object`1269812699        Image content specified directly as base64 data or as a reference via a URL.1270012701        - `type: "image"`1270212703        - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`1270412705          Union type for image source variants.1270612707          - `BetaManagedAgentsBase64ImageSource object`1270812709            Base64-encoded image data.1271012711            - `type: "base64"`1271212713            - `data: string`1271412715              Base64-encoded image data.1271612717              minLength: 11271812719            - `media_type: string`1272012721              MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").1272212723              minLength: 11272412725          - `BetaManagedAgentsURLImageSource object`1272612727            Image referenced by URL.1272812729            - `type: "url"`1273012731            - `url: string`1273212733              URL of the image to fetch.1273412735              minLength: 11273612737          - `BetaManagedAgentsFileImageSource object`1273812739            Image referenced by file ID.1274012741            - `type: "file"`1274212743            - `file_id: string`1274412745              ID of a previously uploaded file.1274612747              minLength: 11274812749      - `BetaManagedAgentsDocumentBlock object`1275012751        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1275212753        - `type: "document"`1275412755        - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`1275612757          Union type for document source variants.1275812759          - `BetaManagedAgentsBase64DocumentSource object`1276012761            Base64-encoded document data.1276212763            - `type: "base64"`1276412765            - `data: string`1276612767              Base64-encoded document data.1276812769              minLength: 11277012771            - `media_type: string`1277212773              MIME type of the document (e.g., "application/pdf").1277412775              minLength: 11277612777          - `BetaManagedAgentsPlainTextDocumentSource object`1277812779            Plain text document content.1278012781            - `type: "text"`1278212783            - `data: string`1278412785              The plain text content.1278612787              minLength: 11278812789            - `media_type: "text/plain"`1279012791              MIME type of the text content. Must be "text/plain".1279212793          - `BetaManagedAgentsURLDocumentSource object`1279412795            Document referenced by URL.1279612797            - `type: "url"`1279812799            - `url: string`1280012801              URL of the document to fetch.1280212803              minLength: 11280412805          - `BetaManagedAgentsFileDocumentSource object`1280612807            Document referenced by file ID.1280812809            - `type: "file"`1281012811            - `file_id: string`1281212813              ID of a previously uploaded file.1281412815              minLength: 11281612817        - `context: optional string or null`1281812819          Additional context about the document for the model.1282012821        - `title: optional string or null`1282212823          The title of the document.1282412825      - `BetaManagedAgentsRedactedBlock object`1282612827        Placeholder for content withheld by Anthropic model policy.1282812829        - `type: "redacted"`1283012831    - `processed_at: optional string or null`1283212833      A timestamp in RFC 3339 format1283412835      format: date-time1283612837  - `BetaManagedAgentsUserInterruptEvent object`1283812839    An interrupt event that pauses agent execution and returns control to the user.1284012841    - `type: "user.interrupt"`1284212843    - `id: string`1284412845      Unique identifier for this event.1284612847    - `processed_at: optional string or null`1284812849      A timestamp in RFC 3339 format1285012851      format: date-time1285212853    - `session_thread_id: optional string or null`1285412855      If absent, interrupts every non-archived thread in a multiagent session (or the primary alone in a single-agent session). If present, interrupts only the named thread.1285612857  - `BetaManagedAgentsUserToolConfirmationEvent object`1285812859    A tool confirmation event that approves or denies a pending tool execution.1286012861    - `type: "user.tool_confirmation"`1286212863    - `id: string`1286412865      Unique identifier for this event.1286612867    - `result: "allow" or "deny"`1286812869      UserToolConfirmationResult enum1287012871      - `"allow"`1287212873      - `"deny"`1287412875    - `tool_use_id: string`1287612877      The id of the `agent.tool_use` or `agent.mcp_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1287812879    - `deny_message: optional string or null`1288012881      Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.1288212883      maxLength: 100001288412885    - `processed_at: optional string or null`1288612887      A timestamp in RFC 3339 format1288812889      format: date-time1289012891    - `session_thread_id: optional string or null`1289212893      Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.1289412895  - `BetaManagedAgentsUserCustomToolResultEvent object`1289612897    Event sent by the client providing the result of a custom tool execution.1289812899    - `type: "user.custom_tool_result"`1290012901    - `id: string`1290212903      Unique identifier for this event.1290412905    - `custom_tool_use_id: string`1290612907      The id of the `agent.custom_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1290812909    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1291012911      The result content returned by the tool.1291212913      - `BetaManagedAgentsTextBlock object`1291412915        Regular text content.1291612917      - `BetaManagedAgentsImageBlock object`1291812919        Image content specified directly as base64 data or as a reference via a URL.1292012921      - `BetaManagedAgentsDocumentBlock object`1292212923        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1292412925      - `BetaManagedAgentsSearchResultBlock object`1292612927        A block containing a web search result.1292812929        - `type: "search_result"`1293012931        - `citations: BetaManagedAgentsSearchResultCitations`1293212933          Citation settings for a search result.1293412935          - `enabled: boolean`1293612937            Whether citations are enabled for this search result.1293812939        - `content: array of BetaManagedAgentsSearchResultContent`1294012941          Array of text content blocks from the search result.1294212943          - `type: "text"`1294412945          - `text: string`1294612947            The text content.1294812949            minLength: 11295012951        - `source: string`1295212953          The URL source of the search result.1295412955          minLength: 11295612957        - `title: string`1295812959          The title of the search result.1296012961          minLength: 11296212963    - `is_error: optional boolean or null`1296412965      Whether the tool execution resulted in an error.1296612967    - `processed_at: optional string or null`1296812969      A timestamp in RFC 3339 format1297012971      format: date-time1297212973    - `session_thread_id: optional string or null`1297412975      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1297612977  - `BetaManagedAgentsAgentCustomToolUseEvent object`1297812979    Event emitted when the agent calls a custom tool. The session goes idle until the client sends a `user.custom_tool_result` event with the result.1298012981    - `type: "agent.custom_tool_use"`1298212983    - `id: string`1298412985      Unique identifier for this event.1298612987    - `input: map[unknown]`1298812989      Input parameters for the tool call.1299012991    - `name: string`1299212993      Name of the custom tool being called.1299412995    - `processed_at: string`1299612997      A timestamp in RFC 3339 format1299812999      format: date-time1300013001    - `session_thread_id: optional string or null`1300213003      When set, this event was cross-posted from a subagent's thread to surface its custom tool use on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.custom_tool_result` by `custom_tool_use_id`, so clients do not send it back.1300413005  - `BetaManagedAgentsAgentMessageEvent object`1300613007    An agent response event in the session conversation.1300813009    - `type: "agent.message"`1301013011    - `id: string`1301213013      Unique identifier for this event.1301413015    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`1301613017      Array of text blocks comprising the agent response.1301813019      - `BetaManagedAgentsTextBlock object`1302013021        Regular text content.1302213023      - `BetaManagedAgentsRedactedBlock object`1302413025        Placeholder for content withheld by Anthropic model policy.1302613027    - `processed_at: string`1302813029      A timestamp in RFC 3339 format1303013031      format: date-time1303213033  - `BetaManagedAgentsAgentThinkingEvent object`1303413035    Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.1303613037    - `type: "agent.thinking"`1303813039    - `id: string`1304013041      Unique identifier for this event.1304213043    - `processed_at: string`1304413045      A timestamp in RFC 3339 format1304613047      format: date-time1304813049  - `BetaManagedAgentsAgentMCPToolUseEvent object`1305013051    Event emitted when the agent invokes a tool provided by an MCP server.1305213053    - `type: "agent.mcp_tool_use"`1305413055    - `id: string`1305613057      Unique identifier for this event.1305813059    - `input: map[unknown]`1306013061      Input parameters for the tool call.1306213063    - `mcp_server_name: string`1306413065      Name of the MCP server providing the tool.1306613067    - `name: string`1306813069      Name of the MCP tool being used.1307013071    - `processed_at: string`1307213073      A timestamp in RFC 3339 format1307413075      format: date-time1307613077    - `evaluated_permission: optional "allow" or "ask" or "deny"`1307813079      AgentEvaluatedPermission enum1308013081      - `"allow"`1308213083      - `"ask"`1308413085      - `"deny"`1308613087    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`1308813089      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.1309013091      - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`1309213093        The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".1309413095        - `type: "always_allow"`1309613097      - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`1309813099        The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".1310013101        - `type: "always_ask"`1310213103      - `BetaManagedAgentsAgentToolEvaluationAuto object`1310413105        The resolved permission_policy was auto: the server judged this invocation individually.1310613107        - `type: "auto"`1310813109        - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`1311013111          The server's per-invocation judgement under the auto permission policy. Its type always equals the event's top-level evaluated_permission. Open union: clients must tolerate unknown variants.1311213113          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`1311413115            The server judged the invocation safe to execute without client approval.1311613117            - `type: "allow"`1311813119          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`1312013121            The server reached no judgement; the invocation is held for client approval.1312213123            - `type: "ask"`1312413125            - `reason_code: string`1312613127              The judgement's grounds in registry-bound terms, for client branching and audit rather than end-user display. Open registry; currently "indeterminate" (no judgement was reached). Clients must tolerate values outside this set.1312813129              maxLength: 641313013131          - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`1313213133            The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.1313413135            - `type: "deny"`1313613137            - `reason_code: string`1313813139              The judgement's grounds in registry-bound terms. Open registry; currently "high_risk" (judged high-risk; the call does not run). Clients must tolerate values outside this set.1314013141              maxLength: 641314213143    - `session_thread_id: optional string or null`1314413145      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` by `tool_use_id`, so clients do not send it back.1314613147  - `BetaManagedAgentsAgentMCPToolResultEvent object`1314813149    Event representing the result of an MCP tool execution.1315013151    - `type: "agent.mcp_tool_result"`1315213153    - `id: string`1315413155      Unique identifier for this event.1315613157    - `mcp_tool_use_id: string`1315813159      The id of the `agent.mcp_tool_use` event this result corresponds to.1316013161    - `processed_at: string`1316213163      A timestamp in RFC 3339 format1316413165      format: date-time1316613167    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1316813169      The result content returned by the tool.1317013171      - `BetaManagedAgentsTextBlock object`1317213173        Regular text content.1317413175      - `BetaManagedAgentsImageBlock object`1317613177        Image content specified directly as base64 data or as a reference via a URL.1317813179      - `BetaManagedAgentsDocumentBlock object`1318013181        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1318213183      - `BetaManagedAgentsSearchResultBlock object`1318413185        A block containing a web search result.1318613187    - `is_error: optional boolean or null`1318813189      Whether the tool execution resulted in an error.1319013191  - `BetaManagedAgentsAgentToolUseEvent object`1319213193    Event emitted when the agent invokes a built-in agent tool.1319413195    - `type: "agent.tool_use"`1319613197    - `id: string`1319813199      Unique identifier for this event.1320013201    - `input: map[unknown]`1320213203      Input parameters for the tool call.1320413205    - `name: string`1320613207      Name of the agent tool being used.1320813209    - `processed_at: string`1321013211      A timestamp in RFC 3339 format1321213213      format: date-time1321413215    - `evaluated_permission: optional "allow" or "ask" or "deny"`1321613217      AgentEvaluatedPermission enum1321813219      - `"allow"`1322013221      - `"ask"`1322213223      - `"deny"`1322413225    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`1322613227      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.1322813229    - `session_thread_id: optional string or null`1323013231      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` or `user.tool_result` by `tool_use_id`, so clients do not send it back.1323213233  - `BetaManagedAgentsAgentToolResultEvent object`1323413235    Event representing the result of an agent tool execution.1323613237    - `type: "agent.tool_result"`1323813239    - `id: string`1324013241      Unique identifier for this event.1324213243    - `processed_at: string`1324413245      A timestamp in RFC 3339 format1324613247      format: date-time1324813249    - `tool_use_id: string`1325013251      The id of the `agent.tool_use` event this result corresponds to.1325213253    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1325413255      The result content returned by the tool.1325613257      - `BetaManagedAgentsTextBlock object`1325813259        Regular text content.1326013261      - `BetaManagedAgentsImageBlock object`1326213263        Image content specified directly as base64 data or as a reference via a URL.1326413265      - `BetaManagedAgentsDocumentBlock object`1326613267        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1326813269      - `BetaManagedAgentsSearchResultBlock object`1327013271        A block containing a web search result.1327213273    - `is_error: optional boolean or null`1327413275      Whether the tool execution resulted in an error.1327613277  - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`1327813279    Delivery event written to the target thread's input stream when an agent-to-agent message arrives.1328013281    - `type: "agent.thread_message_received"`1328213283    - `id: string`1328413285      Unique identifier for this event.1328613287    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1328813289      Message content blocks.1329013291      - `BetaManagedAgentsTextBlock object`1329213293        Regular text content.1329413295      - `BetaManagedAgentsImageBlock object`1329613297        Image content specified directly as base64 data or as a reference via a URL.1329813299      - `BetaManagedAgentsDocumentBlock object`1330013301        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1330213303      - `BetaManagedAgentsRedactedBlock object`1330413305        Placeholder for content withheld by Anthropic model policy.1330613307    - `from_session_thread_id: string`1330813309      Public `sthr_` ID of the thread that sent the message.1331013311    - `processed_at: string`1331213313      A timestamp in RFC 3339 format1331413315      format: date-time1331613317    - `from_agent_name: optional string or null`1331813319      Name of the callable agent this message came from. Absent when received from the primary agent.1332013321  - `BetaManagedAgentsAgentThreadMessageSentEvent object`1332213323    Observability event emitted to the sender's output stream when an agent-to-agent message is sent.1332413325    - `type: "agent.thread_message_sent"`1332613327    - `id: string`1332813329      Unique identifier for this event.1333013331    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1333213333      Message content blocks.1333413335      - `BetaManagedAgentsTextBlock object`1333613337        Regular text content.1333813339      - `BetaManagedAgentsImageBlock object`1334013341        Image content specified directly as base64 data or as a reference via a URL.1334213343      - `BetaManagedAgentsDocumentBlock object`1334413345        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1334613347      - `BetaManagedAgentsRedactedBlock object`1334813349        Placeholder for content withheld by Anthropic model policy.1335013351    - `processed_at: string`1335213353      A timestamp in RFC 3339 format1335413355      format: date-time1335613357    - `to_session_thread_id: string`1335813359      Public `sthr_` ID of the thread the message was sent to.1336013361    - `to_agent_name: optional string or null`1336213363      Name of the callable agent this message was sent to. Absent when sent to the primary agent.1336413365  - `BetaManagedAgentsAgentThreadContextCompactedEvent object`1336613367    Indicates that context compaction (summarization) occurred during the session.1336813369    - `type: "agent.thread_context_compacted"`1337013371    - `id: string`1337213373      Unique identifier for this event.1337413375    - `processed_at: string`1337613377      A timestamp in RFC 3339 format1337813379      format: date-time1338013381  - `BetaManagedAgentsSessionErrorEvent object`1338213383    An error event indicating a problem occurred during session execution.1338413385    - `type: "session.error"`1338613387    - `id: string`1338813389      Unique identifier for this event.1339013391    - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`1339213393      - `BetaManagedAgentsUnknownError object`1339413395        An unknown or unexpected error occurred during session execution. A fallback variant; clients that don't recognize a new error code can match on `retry_status` and `message` alone.1339613397        - `type: "unknown_error"`1339813399        - `message: string`1340013401          Human-readable error description.1340213403        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1340413405          What the client should do next in response to this error.1340613407          - `BetaManagedAgentsRetryStatusRetrying object`1340813409            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1341013411            - `type: "retrying"`1341213413          - `BetaManagedAgentsRetryStatusExhausted object`1341413415            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1341613417            - `type: "exhausted"`1341813419          - `BetaManagedAgentsRetryStatusTerminal object`1342013421            The session encountered a terminal error and will transition to `terminated` state.1342213423            - `type: "terminal"`1342413425      - `BetaManagedAgentsModelOverloadedError object`1342613427        The model is currently overloaded. Emitted after automatic retries are exhausted.1342813429        - `type: "model_overloaded_error"`1343013431        - `message: string`1343213433          Human-readable error description.1343413435        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1343613437          What the client should do next in response to this error.1343813439          - `BetaManagedAgentsRetryStatusRetrying object`1344013441            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1344213443          - `BetaManagedAgentsRetryStatusExhausted object`1344413445            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1344613447          - `BetaManagedAgentsRetryStatusTerminal object`1344813449            The session encountered a terminal error and will transition to `terminated` state.1345013451      - `BetaManagedAgentsModelRateLimitedError object`1345213453        The model request was rate-limited.1345413455        - `type: "model_rate_limited_error"`1345613457        - `message: string`1345813459          Human-readable error description.1346013461        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1346213463          What the client should do next in response to this error.1346413465          - `BetaManagedAgentsRetryStatusRetrying object`1346613467            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1346813469          - `BetaManagedAgentsRetryStatusExhausted object`1347013471            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1347213473          - `BetaManagedAgentsRetryStatusTerminal object`1347413475            The session encountered a terminal error and will transition to `terminated` state.1347613477      - `BetaManagedAgentsModelRequestFailedError object`1347813479        A model request failed for a reason other than overload or rate-limiting.1348013481        - `type: "model_request_failed_error"`1348213483        - `message: string`1348413485          Human-readable error description.1348613487        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1348813489          What the client should do next in response to this error.1349013491          - `BetaManagedAgentsRetryStatusRetrying object`1349213493            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1349413495          - `BetaManagedAgentsRetryStatusExhausted object`1349613497            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1349813499          - `BetaManagedAgentsRetryStatusTerminal object`1350013501            The session encountered a terminal error and will transition to `terminated` state.1350213503      - `BetaManagedAgentsMCPConnectionFailedError object`1350413505        Failed to connect to an MCP server.1350613507        - `type: "mcp_connection_failed_error"`1350813509        - `mcp_server_name: string`1351013511          Name of the MCP server that failed to connect.1351213513        - `message: string`1351413515          Human-readable error description.1351613517        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1351813519          What the client should do next in response to this error.1352013521          - `BetaManagedAgentsRetryStatusRetrying object`1352213523            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1352413525          - `BetaManagedAgentsRetryStatusExhausted object`1352613527            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1352813529          - `BetaManagedAgentsRetryStatusTerminal object`1353013531            The session encountered a terminal error and will transition to `terminated` state.1353213533      - `BetaManagedAgentsMCPAuthenticationFailedError object`1353413535        Authentication to an MCP server failed.1353613537        - `type: "mcp_authentication_failed_error"`1353813539        - `mcp_server_name: string`1354013541          Name of the MCP server that failed authentication.1354213543        - `message: string`1354413545          Human-readable error description.1354613547        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1354813549          What the client should do next in response to this error.1355013551          - `BetaManagedAgentsRetryStatusRetrying object`1355213553            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1355413555          - `BetaManagedAgentsRetryStatusExhausted object`1355613557            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1355813559          - `BetaManagedAgentsRetryStatusTerminal object`1356013561            The session encountered a terminal error and will transition to `terminated` state.1356213563      - `BetaManagedAgentsBillingError object`1356413565        The caller's organization or workspace cannot make model requests — out of credits or spend limit reached. Retrying with the same credentials will not succeed; the caller must resolve the billing state.1356613567        - `type: "billing_error"`1356813569        - `message: string`1357013571          Human-readable error description.1357213573        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1357413575          What the client should do next in response to this error.1357613577          - `BetaManagedAgentsRetryStatusRetrying object`1357813579            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1358013581          - `BetaManagedAgentsRetryStatusExhausted object`1358213583            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1358413585          - `BetaManagedAgentsRetryStatusTerminal object`1358613587            The session encountered a terminal error and will transition to `terminated` state.1358813589      - `BetaManagedAgentsCredentialHostUnreachableError object`1359013591        An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.1359213593        - `type: "credential_host_unreachable_error"`1359413595        - `credential_id: string`1359613597          ID of the affected credential.1359813599        - `message: string`1360013601          Human-readable error description.1360213603        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1360413605          What the client should do next in response to this error.1360613607          - `BetaManagedAgentsRetryStatusRetrying object`1360813609            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1361013611          - `BetaManagedAgentsRetryStatusExhausted object`1361213613            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1361413615          - `BetaManagedAgentsRetryStatusTerminal object`1361613617            The session encountered a terminal error and will transition to `terminated` state.1361813619        - `vault_id: string`1362013621          ID of the vault containing the affected credential.1362213623    - `processed_at: string`1362413625      A timestamp in RFC 3339 format1362613627      format: date-time1362813629  - `BetaManagedAgentsSessionStatusRescheduledEvent object`1363013631    Indicates the session is recovering from an error state and is rescheduled for execution.1363213633    - `type: "session.status_rescheduled"`1363413635    - `id: string`1363613637      Unique identifier for this event.1363813639    - `processed_at: string`1364013641      A timestamp in RFC 3339 format1364213643      format: date-time1364413645  - `BetaManagedAgentsSessionStatusRunningEvent object`1364613647    Indicates the session is actively running and the agent is working.1364813649    - `type: "session.status_running"`1365013651    - `id: string`1365213653      Unique identifier for this event.1365413655    - `processed_at: string`1365613657      A timestamp in RFC 3339 format1365813659      format: date-time1366013661  - `BetaManagedAgentsSessionStatusIdleEvent object`1366213663    Indicates the agent has paused and is awaiting user input.1366413665    - `type: "session.status_idle"`1366613667    - `id: string`1366813669      Unique identifier for this event.1367013671    - `processed_at: string`1367213673      A timestamp in RFC 3339 format1367413675      format: date-time1367613677    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`1367813679      - `BetaManagedAgentsSessionEndTurn object`1368013681        The agent completed its turn naturally and is ready for the next user message.1368213683        - `type: "end_turn"`1368413685      - `BetaManagedAgentsSessionRequiresAction object`1368613687        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.1368813689        - `type: "requires_action"`1369013691        - `event_ids: array of string`1369213693          The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.1369413695      - `BetaManagedAgentsSessionRetriesExhausted object`1369613697        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.1369813699        - `type: "retries_exhausted"`1370013701      - `BetaManagedAgentsSessionBudgetReached object`1370213703        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.1370413705        - `type: "budget_reached"`1370613707  - `BetaManagedAgentsSessionStatusTerminatedEvent object`1370813709    Indicates the session has terminated, either due to an error or completion.1371013711    - `type: "session.status_terminated"`1371213713    - `id: string`1371413715      Unique identifier for this event.1371613717    - `processed_at: string`1371813719      A timestamp in RFC 3339 format1372013721      format: date-time1372213723  - `BetaManagedAgentsSessionThreadCreatedEvent object`1372413725    Emitted when a subagent is spawned as a new thread. Written to the parent thread's output stream so clients observing the session see child creation.1372613727    - `type: "session.thread_created"`1372813729    - `id: string`1373013731      Unique identifier for this event.1373213733    - `agent_name: string`1373413735      Name of the callable agent the thread runs.1373613737    - `processed_at: string`1373813739      A timestamp in RFC 3339 format1374013741      format: date-time1374213743    - `session_thread_id: string`1374413745      Public `sthr_` ID of the newly created thread.1374613747  - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`1374813749    Emitted when an outcome evaluation cycle begins.1375013751    - `type: "span.outcome_evaluation_start"`1375213753    - `id: string`1375413755      Unique identifier for this event.1375613757    - `iteration: number`1375813759      0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.1376013761      format: int321376213763    - `outcome_id: string`1376413765      The `outc_` ID of the outcome being evaluated.1376613767    - `processed_at: string`1376813769      A timestamp in RFC 3339 format1377013771      format: date-time1377213773  - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`1377413775    Emitted when an outcome evaluation cycle completes. Carries the verdict and aggregate token usage. A verdict of `needs_revision` means another evaluation cycle follows; `satisfied`, `max_iterations_reached`, `failed`, or `interrupted` are terminal — no further evaluation cycles follow.1377613777    - `type: "span.outcome_evaluation_end"`1377813779    - `id: string`1378013781      Unique identifier for this event.1378213783    - `explanation: string`1378413785      Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.1378613787    - `iteration: number`1378813789      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.1379013791      format: int321379213793    - `outcome_evaluation_start_id: string`1379413795      The id of the corresponding `span.outcome_evaluation_start` event.1379613797    - `outcome_id: string`1379813799      The `outc_` ID of the outcome being evaluated.1380013801    - `processed_at: string`1380213803      A timestamp in RFC 3339 format1380413805      format: date-time1380613807    - `result: string`1380813809      Evaluation verdict. 'satisfied': criteria met, session goes idle. 'needs_revision': criteria not met, another revision cycle follows. 'max_iterations_reached': evaluation budget exhausted with criteria still unmet — one final acknowledgment turn follows before the session goes idle, but no further evaluation runs. 'failed': grader determined the rubric does not apply to the deliverables. 'interrupted': user sent an interrupt while evaluation was in progress.1381013811    - `usage: BetaManagedAgentsSpanModelUsage`1381213813      Token usage for a single model request.1381413815      - `cache_creation_input_tokens: number`1381613817        Tokens used to create prompt cache in this request.1381813819        format: int321382013821      - `cache_read_input_tokens: number`1382213823        Tokens read from prompt cache in this request.1382413825        format: int321382613827      - `input_tokens: number`1382813829        Input tokens consumed by this request.1383013831        format: int321383213833      - `output_tokens: number`1383413835        Output tokens generated by this request.1383613837        format: int321383813839      - `speed: optional "standard" or "fast" or null`1384013841        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1384213843        - `"standard"`1384413845        - `"fast"`1384613847  - `BetaManagedAgentsSpanModelRequestStartEvent object`1384813849    Emitted when a model request is initiated by the agent.1385013851    - `type: "span.model_request_start"`1385213853    - `id: string`1385413855      Unique identifier for this event.1385613857    - `processed_at: string`1385813859      A timestamp in RFC 3339 format1386013861      format: date-time1386213863  - `BetaManagedAgentsSpanModelRequestEndEvent object`1386413865    Emitted when a model request completes.1386613867    - `type: "span.model_request_end"`1386813869    - `id: string`1387013871      Unique identifier for this event.1387213873    - `is_error: boolean or null`1387413875      Whether the model request resulted in an error.1387613877    - `model_request_start_id: string`1387813879      The id of the corresponding `span.model_request_start` event.1388013881    - `model_usage: BetaManagedAgentsSpanModelUsage`1388213883      Token usage for a single model request.1388413885    - `processed_at: string`1388613887      A timestamp in RFC 3339 format1388813889      format: date-time1389013891  - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`1389213893    Periodic heartbeat emitted while an outcome evaluation cycle is in progress. Distinguishes 'evaluation is actively running' from 'evaluation is stuck' between the corresponding `span.outcome_evaluation_start` and `span.outcome_evaluation_end` events.1389413895    - `type: "span.outcome_evaluation_ongoing"`1389613897    - `id: string`1389813899      Unique identifier for this event.1390013901    - `iteration: number`1390213903      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.1390413905      format: int321390613907    - `outcome_id: string`1390813909      The `outc_` ID of the outcome being evaluated.1391013911    - `processed_at: string`1391213913      A timestamp in RFC 3339 format1391413915      format: date-time1391613917  - `BetaManagedAgentsUserDefineOutcomeEvent object`1391813919    Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.1392013921    - `type: "user.define_outcome"`1392213923    - `id: string`1392413925      Unique identifier for this event.1392613927    - `description: string`1392813929      What the agent should produce. Copied from the input event.1393013931    - `max_iterations: number or null`1393213933      Evaluate-then-revise cycles before giving up. Default 3, max 20.1393413935      format: int321393613937    - `outcome_id: string`1393813939      Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.1394013941    - `processed_at: string`1394213943      A timestamp in RFC 3339 format1394413945      format: date-time1394613947    - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`1394813949      Rubric for grading the quality of an outcome.1395013951      - `BetaManagedAgentsFileRubric object`1395213953        Rubric referenced by a file uploaded via the Files API.1395413955        - `type: "file"`1395613957        - `file_id: string`1395813959          ID of the rubric file.1396013961      - `BetaManagedAgentsTextRubric object`1396213963        Rubric content provided inline as text.1396413965        - `type: "text"`1396613967        - `content: string`1396813969          Rubric content. Plain text or markdown — the grader treats it as freeform text.1397013971  - `BetaManagedAgentsSessionDeletedEvent object`1397213973    Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.1397413975    - `type: "session.deleted"`1397613977    - `id: string`1397813979      Unique identifier for this event.1398013981    - `processed_at: string`1398213983      A timestamp in RFC 3339 format1398413985      format: date-time1398613987  - `BetaManagedAgentsSessionThreadStatusRunningEvent object`1398813989    A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1399013991    - `type: "session.thread_status_running"`1399213993    - `id: string`1399413995      Unique identifier for this event.1399613997    - `agent_name: string`1399813999      Name of the agent the thread runs.1400014001    - `processed_at: string`1400214003      A timestamp in RFC 3339 format1400414005      format: date-time1400614007    - `session_thread_id: string`1400814009      Public sthr_ ID of the thread that started running.1401014011  - `BetaManagedAgentsSessionThreadStatusIdleEvent object`1401214013    A session thread has yielded and is awaiting input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1401414015    - `type: "session.thread_status_idle"`1401614017    - `id: string`1401814019      Unique identifier for this event.1402014021    - `agent_name: string`1402214023      Name of the agent the thread runs.1402414025    - `processed_at: string`1402614027      A timestamp in RFC 3339 format1402814029      format: date-time1403014031    - `session_thread_id: string`1403214033      Public sthr_ ID of the thread that went idle.1403414035    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`1403614037      - `BetaManagedAgentsSessionEndTurn object`1403814039        The agent completed its turn naturally and is ready for the next user message.1404014041      - `BetaManagedAgentsSessionRequiresAction object`1404214043        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.1404414045      - `BetaManagedAgentsSessionRetriesExhausted object`1404614047        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.1404814049      - `BetaManagedAgentsSessionBudgetReached object`1405014051        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.1405214053  - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`1405414055    A session thread has terminated and will accept no further input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1405614057    - `type: "session.thread_status_terminated"`1405814059    - `id: string`1406014061      Unique identifier for this event.1406214063    - `agent_name: string`1406414065      Name of the agent the thread runs.1406614067    - `processed_at: string`1406814069      A timestamp in RFC 3339 format1407014071      format: date-time1407214073    - `session_thread_id: string`1407414075      Public sthr_ ID of the thread that terminated.1407614077  - `BetaManagedAgentsUserToolResultEvent object`1407814079    Event sent by the client providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.1408014081    - `type: "user.tool_result"`1408214083    - `id: string`1408414085      Unique identifier for this event.1408614087    - `tool_use_id: string`1408814089      The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1409014091    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1409214093      The result content returned by the tool.1409414095      - `BetaManagedAgentsTextBlock object`1409614097        Regular text content.1409814099      - `BetaManagedAgentsImageBlock object`1410014101        Image content specified directly as base64 data or as a reference via a URL.1410214103      - `BetaManagedAgentsDocumentBlock object`1410414105        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1410614107      - `BetaManagedAgentsSearchResultBlock object`1410814109        A block containing a web search result.1411014111    - `is_error: optional boolean or null`1411214113      Whether the tool execution resulted in an error.1411414115    - `processed_at: optional string or null`1411614117      A timestamp in RFC 3339 format1411814119      format: date-time1412014121    - `session_thread_id: optional string or null`1412214123      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1412414125  - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`1412614127    A session thread hit a transient error and is retrying automatically. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1412814129    - `type: "session.thread_status_rescheduled"`1413014131    - `id: string`1413214133      Unique identifier for this event.1413414135    - `agent_name: string`1413614137      Name of the agent the thread runs.1413814139    - `processed_at: string`1414014141      A timestamp in RFC 3339 format1414214143      format: date-time1414414145    - `session_thread_id: string`1414614147      Public sthr_ ID of the thread that is retrying.1414814149  - `BetaManagedAgentsSessionUpdatedEvent object`1415014151    Emitted when an UpdateSession request changed at least one field. Carries only the fields that changed; absent fields were not part of the update. The new configuration applies from the next turn.1415214153    - `type: "session.updated"`1415414155    - `id: string`1415614157      Unique identifier for this event.1415814159    - `processed_at: string`1416014161      A timestamp in RFC 3339 format1416214163      format: date-time1416414165    - `agent: optional BetaManagedAgentsSessionAgent or null`1416614167      Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.1416814169      - `type: "agent"`1417014171      - `id: string`1417214173      - `description: string or null`1417414175      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1417614177        - `type: "url"`1417814179        - `name: string`1418014181        - `url: string`1418214183      - `model: BetaManagedAgentsModelConfig`1418414185        Model identifier and configuration.1418614187        - `id: BetaManagedAgentsModel`1418814189          The model that will power your agent.1419014191          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1419214193          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`1419414195            The model that will power your agent.1419614197            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1419814199            - `"claude-fable-5-1"`1420014201              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows1420214203            - `"claude-sonnet-5"`1420414205              High-performance model for coding and agents1420614207            - `"claude-fable-5"`1420814209              Next generation of intelligence for the hardest knowledge work and coding problems1421014211            - `"claude-opus-5"`1421214213              Powerful intelligence for long-running agents and coding1421414215            - `"claude-opus-4-8"`1421614217              Powerful intelligence for long-running agents and coding1421814219            - `"claude-opus-4-7"`1422014221              Powerful intelligence for long-running agents and coding1422214223            - `"claude-opus-4-6"`1422414225              Powerful intelligence for long-running agents and coding1422614227            - `"claude-sonnet-4-6"`1422814229              Best combination of speed and intelligence1423014231            - `"claude-haiku-4-5"`1423214233              Fastest model with near-frontier intelligence1423414235            - `"claude-haiku-4-5-20251001"`1423614237              Fastest model with near-frontier intelligence1423814239            - `"claude-opus-4-5"`1424014241              Powerful intelligence for long-running agents and coding1424214243            - `"claude-opus-4-5-20251101"`1424414245              Powerful intelligence for long-running agents and coding1424614247            - `"claude-sonnet-4-5"`1424814249              High-performance model for agents and coding1425014251            - `"claude-sonnet-4-5-20250929"`1425214253              High-performance model for agents and coding1425414255          - `string`1425614257        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`1425814259          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.1426014261          - `BetaManagedAgentsEffortLow object`1426214263            Low effort. Favors latency over reasoning depth.1426414265            - `type: "low"`1426614267          - `BetaManagedAgentsEffortMedium object`1426814269            Medium effort. Balances latency and reasoning depth.1427014271            - `type: "medium"`1427214273          - `BetaManagedAgentsEffortHigh object`1427414275            High effort. Favors reasoning depth.1427614277            - `type: "high"`1427814279          - `BetaManagedAgentsEffortXhigh object`1428014281            Extra-high effort. Not all models accept this level.1428214283            - `type: "xhigh"`1428414285          - `BetaManagedAgentsEffortMax object`1428614287            Maximum effort. Favors reasoning depth over latency.1428814289            - `type: "max"`1429014291        - `inference_geo: optional string`1429214293          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.1429414295        - `speed: optional "standard" or "fast"`1429614297          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1429814299          - `"standard"`1430014301          - `"fast"`1430214303      - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`1430414305        Resolved coordinator topology with full agent definitions for each roster member.1430614307        - `type: "coordinator"`1430814309        - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`1431014311          Full `agent` definitions the coordinator may spawn as session threads.1431214313          - `BetaManagedAgentsSessionThreadAgent object`1431414315            Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.1431614317            - `type: "agent"`1431814319            - `id: string`1432014321            - `description: string or null`1432214323            - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1432414325              - `type: "url"`1432614327              - `name: string`1432814329              - `url: string`1433014331            - `model: BetaManagedAgentsModelConfig`1433214333              Model identifier and configuration.1433414335            - `name: string`1433614337            - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1433814339              - `BetaManagedAgentsAnthropicSkill object`1434014341                A resolved Anthropic-managed skill.1434214343                - `type: "anthropic"`1434414345                - `skill_id: string`1434614347                - `version: string`1434814349              - `BetaManagedAgentsCustomSkill object`1435014351                A resolved user-created custom skill.1435214353                - `type: "custom"`1435414355                - `skill_id: string`1435614357                - `version: string`1435814359            - `system: string or null`1436014361            - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1436214363              - `BetaManagedAgentsAgentToolset20260401 object`1436414365                - `type: "agent_toolset_20260401"`1436614367                - `configs: array of BetaManagedAgentsAgentToolConfig`1436814369                  - `BetaManagedAgentsBashToolConfig object`1437014371                    Configuration for the bash tool.1437214373                    - `type: "bash"`1437414375                    - `enabled: boolean`1437614377                    - `name: "bash"`1437814379                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1438014381                      Permission policy for tool execution.1438214383                      - `BetaManagedAgentsAlwaysAllowPolicy object`1438414385                        Tool calls are automatically approved without user confirmation.1438614387                        - `type: "always_allow"`1438814389                      - `BetaManagedAgentsAlwaysAskPolicy object`1439014391                        Tool calls require user confirmation before execution.1439214393                        - `type: "always_ask"`1439414395                      - `BetaManagedAgentsAutoPolicy object`1439614397                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1439814399                        - `type: "auto"`1440014401                  - `BetaManagedAgentsEditToolConfig object`1440214403                    Configuration for the edit tool.1440414405                    - `type: "edit"`1440614407                    - `enabled: boolean`1440814409                    - `name: "edit"`1441014411                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1441214413                      Permission policy for tool execution.1441414415                      - `BetaManagedAgentsAlwaysAllowPolicy object`1441614417                        Tool calls are automatically approved without user confirmation.1441814419                      - `BetaManagedAgentsAlwaysAskPolicy object`1442014421                        Tool calls require user confirmation before execution.1442214423                      - `BetaManagedAgentsAutoPolicy object`1442414425                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1442614427                  - `BetaManagedAgentsReadToolConfig object`1442814429                    Configuration for the read tool.1443014431                    - `type: "read"`1443214433                    - `enabled: boolean`1443414435                    - `name: "read"`1443614437                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1443814439                      Permission policy for tool execution.1444014441                      - `BetaManagedAgentsAlwaysAllowPolicy object`1444214443                        Tool calls are automatically approved without user confirmation.1444414445                      - `BetaManagedAgentsAlwaysAskPolicy object`1444614447                        Tool calls require user confirmation before execution.1444814449                      - `BetaManagedAgentsAutoPolicy object`1445014451                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1445214453                  - `BetaManagedAgentsWriteToolConfig object`1445414455                    Configuration for the write tool.1445614457                    - `type: "write"`1445814459                    - `enabled: boolean`1446014461                    - `name: "write"`1446214463                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1446414465                      Permission policy for tool execution.1446614467                      - `BetaManagedAgentsAlwaysAllowPolicy object`1446814469                        Tool calls are automatically approved without user confirmation.1447014471                      - `BetaManagedAgentsAlwaysAskPolicy object`1447214473                        Tool calls require user confirmation before execution.1447414475                      - `BetaManagedAgentsAutoPolicy object`1447614477                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1447814479                  - `BetaManagedAgentsGlobToolConfig object`1448014481                    Configuration for the glob tool.1448214483                    - `type: "glob"`1448414485                    - `enabled: boolean`1448614487                    - `name: "glob"`1448814489                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1449014491                      Permission policy for tool execution.1449214493                      - `BetaManagedAgentsAlwaysAllowPolicy object`1449414495                        Tool calls are automatically approved without user confirmation.1449614497                      - `BetaManagedAgentsAlwaysAskPolicy object`1449814499                        Tool calls require user confirmation before execution.1450014501                      - `BetaManagedAgentsAutoPolicy object`1450214503                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1450414505                  - `BetaManagedAgentsGrepToolConfig object`1450614507                    Configuration for the grep tool.1450814509                    - `type: "grep"`1451014511                    - `enabled: boolean`1451214513                    - `name: "grep"`1451414515                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1451614517                      Permission policy for tool execution.1451814519                      - `BetaManagedAgentsAlwaysAllowPolicy object`1452014521                        Tool calls are automatically approved without user confirmation.1452214523                      - `BetaManagedAgentsAlwaysAskPolicy object`1452414525                        Tool calls require user confirmation before execution.1452614527                      - `BetaManagedAgentsAutoPolicy object`1452814529                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1453014531                  - `BetaManagedAgentsWebFetchToolConfig object`1453214533                    Configuration for the web_fetch tool.1453414535                    - `type: "web_fetch"`1453614537                    - `enabled: boolean`1453814539                    - `name: "web_fetch"`1454014541                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1454214543                      Permission policy for tool execution.1454414545                      - `BetaManagedAgentsAlwaysAllowPolicy object`1454614547                        Tool calls are automatically approved without user confirmation.1454814549                      - `BetaManagedAgentsAlwaysAskPolicy object`1455014551                        Tool calls require user confirmation before execution.1455214553                      - `BetaManagedAgentsAutoPolicy object`1455414555                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1455614557                    - `allowed_domains: optional array of string`1455814559                    - `blocked_domains: optional array of string`1456014561                    - `max_content_tokens: optional number or null`1456214563                      format: int321456414565                  - `BetaManagedAgentsWebSearchToolConfig object`1456614567                    Configuration for the web_search tool.1456814569                    - `type: "web_search"`1457014571                    - `enabled: boolean`1457214573                    - `name: "web_search"`1457414575                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1457614577                      Permission policy for tool execution.1457814579                      - `BetaManagedAgentsAlwaysAllowPolicy object`1458014581                        Tool calls are automatically approved without user confirmation.1458214583                      - `BetaManagedAgentsAlwaysAskPolicy object`1458414585                        Tool calls require user confirmation before execution.1458614587                      - `BetaManagedAgentsAutoPolicy object`1458814589                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1459014591                    - `allowed_domains: optional array of string`1459214593                    - `blocked_domains: optional array of string`1459414595                    - `user_location: optional BetaManagedAgentsUserLocation or null`1459614597                      Approximate user location for search result localization.1459814599                      - `type: "approximate"`1460014601                        Location precision. Only "approximate" is supported.1460214603                      - `city: optional string or null`1460414605                        City name.1460614607                        minLength: 1, maxLength: 2551460814609                      - `country: optional string or null`1461014611                        Two-letter ISO 3166-1 country code, uppercase.1461214613                      - `region: optional string or null`1461414615                        Region or state name.1461614617                        minLength: 1, maxLength: 2551461814619                      - `timezone: optional string or null`1462014621                        IANA timezone identifier, e.g. "America/Los_Angeles".1462214623                        minLength: 1, maxLength: 2551462414625                - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`1462614627                  Resolved default configuration for agent tools.1462814629                  - `enabled: boolean`1463014631                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1463214633                    Permission policy for tool execution.1463414635                    - `BetaManagedAgentsAlwaysAllowPolicy object`1463614637                      Tool calls are automatically approved without user confirmation.1463814639                    - `BetaManagedAgentsAlwaysAskPolicy object`1464014641                      Tool calls require user confirmation before execution.1464214643                    - `BetaManagedAgentsAutoPolicy object`1464414645                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1464614647              - `BetaManagedAgentsMCPToolset object`1464814649                - `type: "mcp_toolset"`1465014651                - `configs: array of BetaManagedAgentsMCPToolConfig`1465214653                  - `enabled: boolean`1465414655                  - `name: string`1465614657                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1465814659                    Permission policy for tool execution.1466014661                    - `BetaManagedAgentsAlwaysAllowPolicy object`1466214663                      Tool calls are automatically approved without user confirmation.1466414665                    - `BetaManagedAgentsAlwaysAskPolicy object`1466614667                      Tool calls require user confirmation before execution.1466814669                    - `BetaManagedAgentsAutoPolicy object`1467014671                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1467214673                - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`1467414675                  Resolved default configuration for all tools from an MCP server.1467614677                  - `enabled: boolean`1467814679                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1468014681                    Permission policy for tool execution.1468214683                    - `BetaManagedAgentsAlwaysAllowPolicy object`1468414685                      Tool calls are automatically approved without user confirmation.1468614687                    - `BetaManagedAgentsAlwaysAskPolicy object`1468814689                      Tool calls require user confirmation before execution.1469014691                    - `BetaManagedAgentsAutoPolicy object`1469214693                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1469414695                - `mcp_server_name: string`1469614697              - `BetaManagedAgentsCustomTool object`1469814699                A custom tool as returned in API responses.1470014701                - `type: "custom"`1470214703                - `description: string`1470414705                - `input_schema: BetaManagedAgentsCustomToolInputSchema`1470614707                  JSON Schema for custom tool input parameters.1470814709                  - `type: "object"`1471014711                  - `properties: optional map[unknown] or null`1471214713                  - `required: optional array of string or null`1471414715                - `name: string`1471614717            - `version: number`1471814719              format: int321472014721          - `BetaManagedAgentsAdvisor object`1472214723            Platform advisor roster entry: a model the session's primary thread may consult mid-turn.1472414725            - `type: "advisor"`1472614727            - `model: string`1472814729              The advisor model id.1473014731      - `name: string`1473214733      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1473414735        - `BetaManagedAgentsAnthropicSkill object`1473614737          A resolved Anthropic-managed skill.1473814739        - `BetaManagedAgentsCustomSkill object`1474014741          A resolved user-created custom skill.1474214743      - `system: string or null`1474414745      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1474614747        - `BetaManagedAgentsAgentToolset20260401 object`1474814749        - `BetaManagedAgentsMCPToolset object`1475014751        - `BetaManagedAgentsCustomTool object`1475214753          A custom tool as returned in API responses.1475414755      - `version: number`1475614757        format: int321475814759    - `budget: optional BetaManagedAgentsBudgetLimit or null`1476014761      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1476214763      - `type: "limit"`1476414765      - `max_list_cost: BetaMonetaryAmount`1476614767        A monetary amount in a specific currency.1476814769        - `amount: string`1477014771          Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.1477214773        - `currency: BetaCurrency`1477414775          Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.1477614777    - `metadata: optional map[string]`1477814779      The session's full metadata bag after the update. Present when the update set non-empty metadata; absent when metadata was unchanged or cleared to empty.1478014781    - `title: optional string or null`1478214783      The session's new title. Present only when the update changed it.1478414785  - `BetaManagedAgentsSystemMessageEvent object`1478614787    A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.1478814789    - `type: "system.message"`1479014791    - `id: string`1479214793      Unique identifier for this event.1479414795    - `content: array of BetaManagedAgentsSystemContentBlock`1479614797      System content blocks. Text-only.1479814799      - `type: "text"`1480014801      - `text: string`1480214803        The text content.1480414805        minLength: 11480614807    - `processed_at: optional string or null`1480814809      A timestamp in RFC 3339 format1481014811      format: date-time1481214813  - `BetaManagedAgentsSessionUsageEvent object`1481414815    Periodic snapshot of the session's cumulative usage and tracked list cost.1481614817    - `type: "session.usage"`1481814819    - `id: string`1482014821      Unique identifier for this event.1482214823    - `processed_at: string`1482414825      A timestamp in RFC 3339 format1482614827      format: date-time1482814829    - `usage: BetaManagedAgentsSessionUsageSnapshot`1483014831      Point-in-time snapshot of a session's cumulative usage.1483214833      - `active_seconds: optional number`1483414835        Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once. This is the duration the session's runtime cost is priced on.1483614837        format: double1483814839      - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`1484014841        Prompt-cache creation token usage broken down by cache lifetime.1484214843        - `ephemeral_1h_input_tokens: optional number`1484414845          Tokens used to create 1-hour ephemeral cache entries.1484614847          format: int321484814849        - `ephemeral_5m_input_tokens: optional number`1485014851          Tokens used to create 5-minute ephemeral cache entries.1485214853          format: int321485414855      - `cache_read_input_tokens: optional number`1485614857        Total tokens read from prompt cache.1485814859        format: int321486014861      - `input_tokens: optional number`1486214863        Total input tokens consumed across all turns.1486414865        format: int321486614867      - `list_cost: optional BetaMonetaryAmount`1486814869        A monetary amount in a specific currency.1487014871      - `output_tokens: optional number`1487214873        Total output tokens generated across all turns.1487414875        format: int321487614877      - `server_tool_use: optional BetaManagedAgentsServerToolUsage`1487814879        Cumulative count of server-executed tool invocations, broken down by tool.1488014881        - `web_fetch_requests: optional number`1488214883          Number of server-executed web fetch requests.1488414885          format: int321488614887        - `web_search_requests: optional number`1488814889          Number of server-executed web search requests.1489014891          format: int321489214893    - `budget: optional BetaManagedAgentsBudgetLimit or null`1489414895      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1489614897- `next_page: optional string or null`1489814899  Opaque cursor for the next page. Null when no more results.1490014901#### Example1490214903```bash14904curl https://api.anthropic.com/v1/sessions/$SESSION_ID/events \14905    -H 'anthropic-version: 2023-06-01' \14906    -H 'anthropic-beta: managed-agents-2026-04-01' \14907    -H "X-Api-Key: $ANTHROPIC_API_KEY"14908```1490914910##### Response (200)1491114912```json14913{14914  "data": [14915    {14916      "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",14917      "content": [14918        {14919          "text": "Where is my order #1234?",14920          "type": "text"14921        }14922      ],14923      "type": "user.message",14924      "processed_at": "2026-03-15T10:00:00Z"14925    },14926    {14927      "id": "sevt_011CZkZHPq1jCdq5lbRTjiVnz",14928      "content": [14929        {14930          "text": "Let me look up order #1234 for you.",14931          "type": "text"14932        }14933      ],14934      "processed_at": "2026-03-15T10:00:00Z",14935      "type": "agent.message"14936    }14937  ],14938  "next_page": "page_MjAyNS0wNS0xNFQwMDowMDowMFo="14939}14940```1494114942### Send Events1494314944**POST** `/v1/sessions/{session_id}/events`1494514946Send Events1494714948#### Path parameters1494914950- `session_id: string`1495114952#### Headers1495314954- `"anthropic-beta": optional array of AnthropicBeta`1495514956  Optional header to specify the beta version(s) you want to use.1495714958  - `string`1495914960  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1496114962    - `"message-batches-2024-09-24"`1496314964    - `"prompt-caching-2024-07-31"`1496514966    - `"computer-use-2024-10-22"`1496714968    - `"computer-use-2025-01-24"`1496914970    - `"pdfs-2024-09-25"`1497114972    - `"token-counting-2024-11-01"`1497314974    - `"token-efficient-tools-2025-02-19"`1497514976    - `"output-128k-2025-02-19"`1497714978    - `"files-api-2025-04-14"`1497914980    - `"mcp-client-2025-04-04"`1498114982    - `"mcp-client-2025-11-20"`1498314984    - `"dev-full-thinking-2025-05-14"`1498514986    - `"interleaved-thinking-2025-05-14"`1498714988    - `"code-execution-2025-05-22"`1498914990    - `"extended-cache-ttl-2025-04-11"`1499114992    - `"context-1m-2025-08-07"`1499314994    - `"context-management-2025-06-27"`1499514996    - `"model-context-window-exceeded-2025-08-26"`1499714998    - `"skills-2025-10-02"`1499915000    - `"fast-mode-2026-02-01"`1500115002    - `"output-300k-2026-03-24"`1500315004    - `"user-profiles-2026-03-24"`1500515006    - `"user-profiles-2026-08-18"`1500715008    - `"user-profiles-2026-09-04"`1500915010    - `"advisor-tool-2026-03-01"`1501115012    - `"managed-agents-2026-04-01"`1501315014    - `"cache-diagnosis-2026-04-07"`1501515016    - `"dreaming-2026-04-21"`1501715018    - `"thinking-token-count-2026-05-13"`1501915020    - `"server-side-fallback-2026-06-01"`1502115022    - `"server-side-fallback-2026-07-01"`1502315024    - `"fallback-credit-2026-06-01"`1502515026    - `"fallback-credit-2026-07-01"`1502715028    - `"agent-memory-2026-07-22"`1502915030    - `"mid-conversation-tool-changes-2026-07-01"`1503115032    - `"compact-2026-01-12"`1503315034    - `"computer-use-2025-11-24"`1503515036    - `"mcp-tunnels-2026-06-22"`1503715038    - `"structured-outputs-2025-11-13"`1503915040    - `"task-budgets-2026-03-13"`1504115042    - `"thinking-display-updates-2026-08-18"`1504315044    - `"ce-user-management-2026-07-13"`1504515046    - `"mid-conversation-output-config-2026-07-01"`1504715048    - `"thinking-binding-controls-2026-08-01"`1504915050    - `"mid-conversation-system-clear-at-2026-08-21"`1505115052    - `"compact-2026-09-04"`1505315054- `"anthropic-workspace-id": optional string`1505515056#### Body parameters1505715058- `events: array of BetaManagedAgentsEventParams`1505915060  Events to send to the `session`.1506115062  - `BetaManagedAgentsUserMessageEventParams object`1506315064    Parameters for sending a user message to the session.1506515066    - `type: "user.message"`1506715068    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1506915070      Array of content blocks for the user message.1507115072      - `BetaManagedAgentsTextBlock object`1507315074        Regular text content.1507515076        - `type: "text"`1507715078        - `text: string`1507915080          The text content.1508115082          minLength: 11508315084      - `BetaManagedAgentsImageBlock object`1508515086        Image content specified directly as base64 data or as a reference via a URL.1508715088        - `type: "image"`1508915090        - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`1509115092          Union type for image source variants.1509315094          - `BetaManagedAgentsBase64ImageSource object`1509515096            Base64-encoded image data.1509715098            - `type: "base64"`1509915100            - `data: string`1510115102              Base64-encoded image data.1510315104              minLength: 11510515106            - `media_type: string`1510715108              MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").1510915110              minLength: 11511115112          - `BetaManagedAgentsURLImageSource object`1511315114            Image referenced by URL.1511515116            - `type: "url"`1511715118            - `url: string`1511915120              URL of the image to fetch.1512115122              minLength: 11512315124          - `BetaManagedAgentsFileImageSource object`1512515126            Image referenced by file ID.1512715128            - `type: "file"`1512915130            - `file_id: string`1513115132              ID of a previously uploaded file.1513315134              minLength: 11513515136      - `BetaManagedAgentsDocumentBlock object`1513715138        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1513915140        - `type: "document"`1514115142        - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`1514315144          Union type for document source variants.1514515146          - `BetaManagedAgentsBase64DocumentSource object`1514715148            Base64-encoded document data.1514915150            - `type: "base64"`1515115152            - `data: string`1515315154              Base64-encoded document data.1515515156              minLength: 11515715158            - `media_type: string`1515915160              MIME type of the document (e.g., "application/pdf").1516115162              minLength: 11516315164          - `BetaManagedAgentsPlainTextDocumentSource object`1516515166            Plain text document content.1516715168            - `type: "text"`1516915170            - `data: string`1517115172              The plain text content.1517315174              minLength: 11517515176            - `media_type: "text/plain"`1517715178              MIME type of the text content. Must be "text/plain".1517915180          - `BetaManagedAgentsURLDocumentSource object`1518115182            Document referenced by URL.1518315184            - `type: "url"`1518515186            - `url: string`1518715188              URL of the document to fetch.1518915190              minLength: 11519115192          - `BetaManagedAgentsFileDocumentSource object`1519315194            Document referenced by file ID.1519515196            - `type: "file"`1519715198            - `file_id: string`1519915200              ID of a previously uploaded file.1520115202              minLength: 11520315204        - `context: optional string or null`1520515206          Additional context about the document for the model.1520715208        - `title: optional string or null`1520915210          The title of the document.1521115212      - `BetaManagedAgentsRedactedBlock object`1521315214        Placeholder for content withheld by Anthropic model policy.1521515216        - `type: "redacted"`1521715218  - `BetaManagedAgentsUserInterruptEventParams object`1521915220    Parameters for sending an interrupt to pause the agent.1522115222    - `type: "user.interrupt"`1522315224    - `session_thread_id: optional string or null`1522515226      If absent, interrupts every non-archived thread in a multiagent session (or the primary alone in a single-agent session). If present, interrupts only the named thread.1522715228  - `BetaManagedAgentsUserToolConfirmationEventParams object`1522915230    Parameters for confirming or denying a tool execution request.1523115232    - `type: "user.tool_confirmation"`1523315234    - `result: "allow" or "deny"`1523515236      UserToolConfirmationResult enum1523715238      - `"allow"`1523915240      - `"deny"`1524115242    - `tool_use_id: string`1524315244      The id of the `agent.tool_use` or `agent.mcp_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1524515246      minLength: 1, maxLength: 1281524715248    - `deny_message: optional string or null`1524915250      Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.1525115252      maxLength: 100001525315254  - `BetaManagedAgentsUserCustomToolResultEventParams object`1525515256    Parameters for providing the result of a custom tool execution.1525715258    - `type: "user.custom_tool_result"`1525915260    - `custom_tool_use_id: string`1526115262      The id of the `agent.custom_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1526315264      minLength: 1, maxLength: 1281526515266    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1526715268      The result content returned by the tool.1526915270      - `BetaManagedAgentsTextBlock object`1527115272        Regular text content.1527315274      - `BetaManagedAgentsImageBlock object`1527515276        Image content specified directly as base64 data or as a reference via a URL.1527715278      - `BetaManagedAgentsDocumentBlock object`1527915280        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1528115282      - `BetaManagedAgentsSearchResultBlock object`1528315284        A block containing a web search result.1528515286        - `type: "search_result"`1528715288        - `citations: BetaManagedAgentsSearchResultCitations`1528915290          Citation settings for a search result.1529115292          - `enabled: boolean`1529315294            Whether citations are enabled for this search result.1529515296        - `content: array of BetaManagedAgentsSearchResultContent`1529715298          Array of text content blocks from the search result.1529915300          - `type: "text"`1530115302          - `text: string`1530315304            The text content.1530515306            minLength: 11530715308        - `source: string`1530915310          The URL source of the search result.1531115312          minLength: 11531315314        - `title: string`1531515316          The title of the search result.1531715318          minLength: 11531915320    - `is_error: optional boolean or null`1532115322      Whether the tool execution resulted in an error.1532315324  - `BetaManagedAgentsUserDefineOutcomeEventParams object`1532515326    Parameters for defining an outcome the agent should work toward. The agent begins work on receipt.1532715328    - `type: "user.define_outcome"`1532915330    - `description: string`1533115332      What the agent should produce. This is the task specification.1533315334    - `rubric: BetaManagedAgentsFileRubricParams or BetaManagedAgentsTextRubricParams`1533515336      Rubric for grading the quality of an outcome.1533715338      - `BetaManagedAgentsFileRubricParams object`1533915340        Rubric referenced by a file uploaded via the Files API.1534115342        - `type: "file"`1534315344        - `file_id: string`1534515346          ID of the rubric file.1534715348      - `BetaManagedAgentsTextRubricParams object`1534915350        Rubric content provided inline as text.1535115352        - `type: "text"`1535315354        - `content: string`1535515356          Rubric content. Plain text or markdown — the grader treats it as freeform text. Maximum 262144 characters.1535715358          maxLength: 2621441535915360    - `max_iterations: optional number or null`1536115362      Eval→revision cycles before giving up. Default 3, max 20.1536315364      format: int321536515366  - `BetaManagedAgentsUserToolResultEventParams object`1536715368    Parameters for providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.1536915370    - `type: "user.tool_result"`1537115372    - `tool_use_id: string`1537315374      The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1537515376      minLength: 1, maxLength: 1281537715378    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1537915380      The result content returned by the tool.1538115382      - `BetaManagedAgentsTextBlock object`1538315384        Regular text content.1538515386      - `BetaManagedAgentsImageBlock object`1538715388        Image content specified directly as base64 data or as a reference via a URL.1538915390      - `BetaManagedAgentsDocumentBlock object`1539115392        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1539315394      - `BetaManagedAgentsSearchResultBlock object`1539515396        A block containing a web search result.1539715398    - `is_error: optional boolean or null`1539915400      Whether the tool execution resulted in an error.1540115402  - `BetaManagedAgentsSystemMessageEventParams object`1540315404    Privileged context for the accompanying turn and all subsequent turns, appended to the session's system context as a `role: "system"` turn rather than replacing the top-level system prompt. At most one per request: it must be the final event and immediately follow the `user.message`, `user.tool_result`, or `user.custom_tool_result` it accompanies. Only supported on models that accept mid-conversation system messages.1540515406    - `type: "system.message"`1540715408    - `content: array of BetaManagedAgentsSystemContentBlock`1540915410      System content blocks to append. Text-only.1541115412      - `type: "text"`1541315414      - `text: string`1541515416        The text content.1541715418        minLength: 11541915420#### Returns1542115422- `BetaManagedAgentsSendSessionEvents object`1542315424  Events that were successfully sent to the session.1542515426  - `data: optional array of BetaManagedAgentsUserMessageEvent or BetaManagedAgentsUserInterruptEvent or BetaManagedAgentsUserToolConfirmationEvent or 4 more`1542715428    Sent events1542915430    - `BetaManagedAgentsUserMessageEvent object`1543115432      A user message event in the session conversation.1543315434      - `type: "user.message"`1543515436      - `id: string`1543715438        Unique identifier for this event.1543915440      - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1544115442        Array of content blocks comprising the user message.1544315444        - `BetaManagedAgentsTextBlock object`1544515446          Regular text content.1544715448          - `type: "text"`1544915450          - `text: string`1545115452            The text content.1545315454            minLength: 11545515456        - `BetaManagedAgentsImageBlock object`1545715458          Image content specified directly as base64 data or as a reference via a URL.1545915460          - `type: "image"`1546115462          - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`1546315464            Union type for image source variants.1546515466            - `BetaManagedAgentsBase64ImageSource object`1546715468              Base64-encoded image data.1546915470              - `type: "base64"`1547115472              - `data: string`1547315474                Base64-encoded image data.1547515476                minLength: 11547715478              - `media_type: string`1547915480                MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").1548115482                minLength: 11548315484            - `BetaManagedAgentsURLImageSource object`1548515486              Image referenced by URL.1548715488              - `type: "url"`1548915490              - `url: string`1549115492                URL of the image to fetch.1549315494                minLength: 11549515496            - `BetaManagedAgentsFileImageSource object`1549715498              Image referenced by file ID.1549915500              - `type: "file"`1550115502              - `file_id: string`1550315504                ID of a previously uploaded file.1550515506                minLength: 11550715508        - `BetaManagedAgentsDocumentBlock object`1550915510          Document content, either specified directly as base64 data, as text, or as a reference via a URL.1551115512          - `type: "document"`1551315514          - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`1551515516            Union type for document source variants.1551715518            - `BetaManagedAgentsBase64DocumentSource object`1551915520              Base64-encoded document data.1552115522              - `type: "base64"`1552315524              - `data: string`1552515526                Base64-encoded document data.1552715528                minLength: 11552915530              - `media_type: string`1553115532                MIME type of the document (e.g., "application/pdf").1553315534                minLength: 11553515536            - `BetaManagedAgentsPlainTextDocumentSource object`1553715538              Plain text document content.1553915540              - `type: "text"`1554115542              - `data: string`1554315544                The plain text content.1554515546                minLength: 11554715548              - `media_type: "text/plain"`1554915550                MIME type of the text content. Must be "text/plain".1555115552            - `BetaManagedAgentsURLDocumentSource object`1555315554              Document referenced by URL.1555515556              - `type: "url"`1555715558              - `url: string`1555915560                URL of the document to fetch.1556115562                minLength: 11556315564            - `BetaManagedAgentsFileDocumentSource object`1556515566              Document referenced by file ID.1556715568              - `type: "file"`1556915570              - `file_id: string`1557115572                ID of a previously uploaded file.1557315574                minLength: 11557515576          - `context: optional string or null`1557715578            Additional context about the document for the model.1557915580          - `title: optional string or null`1558115582            The title of the document.1558315584        - `BetaManagedAgentsRedactedBlock object`1558515586          Placeholder for content withheld by Anthropic model policy.1558715588          - `type: "redacted"`1558915590      - `processed_at: optional string or null`1559115592        A timestamp in RFC 3339 format1559315594        format: date-time1559515596    - `BetaManagedAgentsUserInterruptEvent object`1559715598      An interrupt event that pauses agent execution and returns control to the user.1559915600      - `type: "user.interrupt"`1560115602      - `id: string`1560315604        Unique identifier for this event.1560515606      - `processed_at: optional string or null`1560715608        A timestamp in RFC 3339 format1560915610        format: date-time1561115612      - `session_thread_id: optional string or null`1561315614        If absent, interrupts every non-archived thread in a multiagent session (or the primary alone in a single-agent session). If present, interrupts only the named thread.1561515616    - `BetaManagedAgentsUserToolConfirmationEvent object`1561715618      A tool confirmation event that approves or denies a pending tool execution.1561915620      - `type: "user.tool_confirmation"`1562115622      - `id: string`1562315624        Unique identifier for this event.1562515626      - `result: "allow" or "deny"`1562715628        UserToolConfirmationResult enum1562915630        - `"allow"`1563115632        - `"deny"`1563315634      - `tool_use_id: string`1563515636        The id of the `agent.tool_use` or `agent.mcp_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1563715638      - `deny_message: optional string or null`1563915640        Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.1564115642        maxLength: 100001564315644      - `processed_at: optional string or null`1564515646        A timestamp in RFC 3339 format1564715648        format: date-time1564915650      - `session_thread_id: optional string or null`1565115652        Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.1565315654    - `BetaManagedAgentsUserCustomToolResultEvent object`1565515656      Event sent by the client providing the result of a custom tool execution.1565715658      - `type: "user.custom_tool_result"`1565915660      - `id: string`1566115662        Unique identifier for this event.1566315664      - `custom_tool_use_id: string`1566515666        The id of the `agent.custom_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1566715668      - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1566915670        The result content returned by the tool.1567115672        - `BetaManagedAgentsTextBlock object`1567315674          Regular text content.1567515676        - `BetaManagedAgentsImageBlock object`1567715678          Image content specified directly as base64 data or as a reference via a URL.1567915680        - `BetaManagedAgentsDocumentBlock object`1568115682          Document content, either specified directly as base64 data, as text, or as a reference via a URL.1568315684        - `BetaManagedAgentsSearchResultBlock object`1568515686          A block containing a web search result.1568715688          - `type: "search_result"`1568915690          - `citations: BetaManagedAgentsSearchResultCitations`1569115692            Citation settings for a search result.1569315694            - `enabled: boolean`1569515696              Whether citations are enabled for this search result.1569715698          - `content: array of BetaManagedAgentsSearchResultContent`1569915700            Array of text content blocks from the search result.1570115702            - `type: "text"`1570315704            - `text: string`1570515706              The text content.1570715708              minLength: 11570915710          - `source: string`1571115712            The URL source of the search result.1571315714            minLength: 11571515716          - `title: string`1571715718            The title of the search result.1571915720            minLength: 11572115722      - `is_error: optional boolean or null`1572315724        Whether the tool execution resulted in an error.1572515726      - `processed_at: optional string or null`1572715728        A timestamp in RFC 3339 format1572915730        format: date-time1573115732      - `session_thread_id: optional string or null`1573315734        Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1573515736    - `BetaManagedAgentsUserDefineOutcomeEvent object`1573715738      Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.1573915740      - `type: "user.define_outcome"`1574115742      - `id: string`1574315744        Unique identifier for this event.1574515746      - `description: string`1574715748        What the agent should produce. Copied from the input event.1574915750      - `max_iterations: number or null`1575115752        Evaluate-then-revise cycles before giving up. Default 3, max 20.1575315754        format: int321575515756      - `outcome_id: string`1575715758        Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.1575915760      - `processed_at: string`1576115762        A timestamp in RFC 3339 format1576315764        format: date-time1576515766      - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`1576715768        Rubric for grading the quality of an outcome.1576915770        - `BetaManagedAgentsFileRubric object`1577115772          Rubric referenced by a file uploaded via the Files API.1577315774          - `type: "file"`1577515776          - `file_id: string`1577715778            ID of the rubric file.1577915780        - `BetaManagedAgentsTextRubric object`1578115782          Rubric content provided inline as text.1578315784          - `type: "text"`1578515786          - `content: string`1578715788            Rubric content. Plain text or markdown — the grader treats it as freeform text.1578915790    - `BetaManagedAgentsUserToolResultEvent object`1579115792      Event sent by the client providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.1579315794      - `type: "user.tool_result"`1579515796      - `id: string`1579715798        Unique identifier for this event.1579915800      - `tool_use_id: string`1580115802        The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1580315804      - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1580515806        The result content returned by the tool.1580715808        - `BetaManagedAgentsTextBlock object`1580915810          Regular text content.1581115812        - `BetaManagedAgentsImageBlock object`1581315814          Image content specified directly as base64 data or as a reference via a URL.1581515816        - `BetaManagedAgentsDocumentBlock object`1581715818          Document content, either specified directly as base64 data, as text, or as a reference via a URL.1581915820        - `BetaManagedAgentsSearchResultBlock object`1582115822          A block containing a web search result.1582315824      - `is_error: optional boolean or null`1582515826        Whether the tool execution resulted in an error.1582715828      - `processed_at: optional string or null`1582915830        A timestamp in RFC 3339 format1583115832        format: date-time1583315834      - `session_thread_id: optional string or null`1583515836        Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1583715838    - `BetaManagedAgentsSystemMessageEvent object`1583915840      A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.1584115842      - `type: "system.message"`1584315844      - `id: string`1584515846        Unique identifier for this event.1584715848      - `content: array of BetaManagedAgentsSystemContentBlock`1584915850        System content blocks. Text-only.1585115852        - `type: "text"`1585315854        - `text: string`1585515856          The text content.1585715858          minLength: 11585915860      - `processed_at: optional string or null`1586115862        A timestamp in RFC 3339 format1586315864        format: date-time1586515866#### Example1586715868```bash15869curl https://api.anthropic.com/v1/sessions/$SESSION_ID/events \15870    -H 'Content-Type: application/json' \15871    -H 'anthropic-version: 2023-06-01' \15872    -H 'anthropic-beta: managed-agents-2026-04-01' \15873    -H "X-Api-Key: $ANTHROPIC_API_KEY" \15874    -d '{15875          "events": [15876            {15877              "content": [15878                {15879                  "text": "Where is my order #1234?",15880                  "type": "text"15881                }15882              ],15883              "type": "user.message"15884            }15885          ]15886        }'15887```1588815889##### Response (200)1589015891```json15892{15893  "data": [15894    {15895      "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",15896      "content": [15897        {15898          "text": "Where is my order #1234?",15899          "type": "text"15900        }15901      ],15902      "type": "user.message",15903      "processed_at": "2026-03-15T10:00:00Z"15904    }15905  ]15906}15907```1590815909### Stream Events1591015911**GET** `/v1/sessions/{session_id}/events/stream`1591215913Stream Events1591415915#### Path parameters1591615917- `session_id: string`1591815919#### Query parameters1592015921- `event_deltas: optional array of BetaManagedAgentsDeltaType`1592215923  When set, this connection also receives streaming deltas (`event_start`, `event_delta`) while an event is being produced, before the event itself arrives. Deltas are best-effort; when the final event is produced it carries the complete content. A model request that ends early (an error or interrupt) produces no final event — its terminal `span.model_request_end` closes the preview. Accepts one or more event types to preview and may be repeated: `agent.message` streams `content_delta` fragments; `agent.thinking` is start-only — a signal that the agent has begun extended thinking, concluded by the `agent.thinking` event itself. Only previews of the requested event types are sent.1592415925  - `"agent.message"`1592615927  - `"agent.thinking"`1592815929#### Headers1593015931- `"anthropic-beta": optional array of AnthropicBeta`1593215933  Optional header to specify the beta version(s) you want to use.1593415935  - `string`1593615937  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1593815939    - `"message-batches-2024-09-24"`1594015941    - `"prompt-caching-2024-07-31"`1594215943    - `"computer-use-2024-10-22"`1594415945    - `"computer-use-2025-01-24"`1594615947    - `"pdfs-2024-09-25"`1594815949    - `"token-counting-2024-11-01"`1595015951    - `"token-efficient-tools-2025-02-19"`1595215953    - `"output-128k-2025-02-19"`1595415955    - `"files-api-2025-04-14"`1595615957    - `"mcp-client-2025-04-04"`1595815959    - `"mcp-client-2025-11-20"`1596015961    - `"dev-full-thinking-2025-05-14"`1596215963    - `"interleaved-thinking-2025-05-14"`1596415965    - `"code-execution-2025-05-22"`1596615967    - `"extended-cache-ttl-2025-04-11"`1596815969    - `"context-1m-2025-08-07"`1597015971    - `"context-management-2025-06-27"`1597215973    - `"model-context-window-exceeded-2025-08-26"`1597415975    - `"skills-2025-10-02"`1597615977    - `"fast-mode-2026-02-01"`1597815979    - `"output-300k-2026-03-24"`1598015981    - `"user-profiles-2026-03-24"`1598215983    - `"user-profiles-2026-08-18"`1598415985    - `"user-profiles-2026-09-04"`1598615987    - `"advisor-tool-2026-03-01"`1598815989    - `"managed-agents-2026-04-01"`1599015991    - `"cache-diagnosis-2026-04-07"`1599215993    - `"dreaming-2026-04-21"`1599415995    - `"thinking-token-count-2026-05-13"`1599615997    - `"server-side-fallback-2026-06-01"`1599815999    - `"server-side-fallback-2026-07-01"`1600016001    - `"fallback-credit-2026-06-01"`1600216003    - `"fallback-credit-2026-07-01"`1600416005    - `"agent-memory-2026-07-22"`1600616007    - `"mid-conversation-tool-changes-2026-07-01"`1600816009    - `"compact-2026-01-12"`1601016011    - `"computer-use-2025-11-24"`1601216013    - `"mcp-tunnels-2026-06-22"`1601416015    - `"structured-outputs-2025-11-13"`1601616017    - `"task-budgets-2026-03-13"`1601816019    - `"thinking-display-updates-2026-08-18"`1602016021    - `"ce-user-management-2026-07-13"`1602216023    - `"mid-conversation-output-config-2026-07-01"`1602416025    - `"thinking-binding-controls-2026-08-01"`1602616027    - `"mid-conversation-system-clear-at-2026-08-21"`1602816029    - `"compact-2026-09-04"`1603016031- `"anthropic-workspace-id": optional string`1603216033#### Returns1603416035- `BetaManagedAgentsStreamSessionEvents = BetaManagedAgentsUserMessageEvent or BetaManagedAgentsUserInterruptEvent or BetaManagedAgentsUserToolConfirmationEvent or 34 more`1603616037  Server-sent event in the session stream.1603816039  - `BetaManagedAgentsUserMessageEvent object`1604016041    A user message event in the session conversation.1604216043    - `type: "user.message"`1604416045    - `id: string`1604616047      Unique identifier for this event.1604816049    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1605016051      Array of content blocks comprising the user message.1605216053      - `BetaManagedAgentsTextBlock object`1605416055        Regular text content.1605616057        - `type: "text"`1605816059        - `text: string`1606016061          The text content.1606216063          minLength: 11606416065      - `BetaManagedAgentsImageBlock object`1606616067        Image content specified directly as base64 data or as a reference via a URL.1606816069        - `type: "image"`1607016071        - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`1607216073          Union type for image source variants.1607416075          - `BetaManagedAgentsBase64ImageSource object`1607616077            Base64-encoded image data.1607816079            - `type: "base64"`1608016081            - `data: string`1608216083              Base64-encoded image data.1608416085              minLength: 11608616087            - `media_type: string`1608816089              MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").1609016091              minLength: 11609216093          - `BetaManagedAgentsURLImageSource object`1609416095            Image referenced by URL.1609616097            - `type: "url"`1609816099            - `url: string`1610016101              URL of the image to fetch.1610216103              minLength: 11610416105          - `BetaManagedAgentsFileImageSource object`1610616107            Image referenced by file ID.1610816109            - `type: "file"`1611016111            - `file_id: string`1611216113              ID of a previously uploaded file.1611416115              minLength: 11611616117      - `BetaManagedAgentsDocumentBlock object`1611816119        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1612016121        - `type: "document"`1612216123        - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`1612416125          Union type for document source variants.1612616127          - `BetaManagedAgentsBase64DocumentSource object`1612816129            Base64-encoded document data.1613016131            - `type: "base64"`1613216133            - `data: string`1613416135              Base64-encoded document data.1613616137              minLength: 11613816139            - `media_type: string`1614016141              MIME type of the document (e.g., "application/pdf").1614216143              minLength: 11614416145          - `BetaManagedAgentsPlainTextDocumentSource object`1614616147            Plain text document content.1614816149            - `type: "text"`1615016151            - `data: string`1615216153              The plain text content.1615416155              minLength: 11615616157            - `media_type: "text/plain"`1615816159              MIME type of the text content. Must be "text/plain".1616016161          - `BetaManagedAgentsURLDocumentSource object`1616216163            Document referenced by URL.1616416165            - `type: "url"`1616616167            - `url: string`1616816169              URL of the document to fetch.1617016171              minLength: 11617216173          - `BetaManagedAgentsFileDocumentSource object`1617416175            Document referenced by file ID.1617616177            - `type: "file"`1617816179            - `file_id: string`1618016181              ID of a previously uploaded file.1618216183              minLength: 11618416185        - `context: optional string or null`1618616187          Additional context about the document for the model.1618816189        - `title: optional string or null`1619016191          The title of the document.1619216193      - `BetaManagedAgentsRedactedBlock object`1619416195        Placeholder for content withheld by Anthropic model policy.1619616197        - `type: "redacted"`1619816199    - `processed_at: optional string or null`1620016201      A timestamp in RFC 3339 format1620216203      format: date-time1620416205  - `BetaManagedAgentsUserInterruptEvent object`1620616207    An interrupt event that pauses agent execution and returns control to the user.1620816209    - `type: "user.interrupt"`1621016211    - `id: string`1621216213      Unique identifier for this event.1621416215    - `processed_at: optional string or null`1621616217      A timestamp in RFC 3339 format1621816219      format: date-time1622016221    - `session_thread_id: optional string or null`1622216223      If absent, interrupts every non-archived thread in a multiagent session (or the primary alone in a single-agent session). If present, interrupts only the named thread.1622416225  - `BetaManagedAgentsUserToolConfirmationEvent object`1622616227    A tool confirmation event that approves or denies a pending tool execution.1622816229    - `type: "user.tool_confirmation"`1623016231    - `id: string`1623216233      Unique identifier for this event.1623416235    - `result: "allow" or "deny"`1623616237      UserToolConfirmationResult enum1623816239      - `"allow"`1624016241      - `"deny"`1624216243    - `tool_use_id: string`1624416245      The id of the `agent.tool_use` or `agent.mcp_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1624616247    - `deny_message: optional string or null`1624816249      Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.1625016251      maxLength: 100001625216253    - `processed_at: optional string or null`1625416255      A timestamp in RFC 3339 format1625616257      format: date-time1625816259    - `session_thread_id: optional string or null`1626016261      Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.1626216263  - `BetaManagedAgentsUserCustomToolResultEvent object`1626416265    Event sent by the client providing the result of a custom tool execution.1626616267    - `type: "user.custom_tool_result"`1626816269    - `id: string`1627016271      Unique identifier for this event.1627216273    - `custom_tool_use_id: string`1627416275      The id of the `agent.custom_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1627616277    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1627816279      The result content returned by the tool.1628016281      - `BetaManagedAgentsTextBlock object`1628216283        Regular text content.1628416285      - `BetaManagedAgentsImageBlock object`1628616287        Image content specified directly as base64 data or as a reference via a URL.1628816289      - `BetaManagedAgentsDocumentBlock object`1629016291        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1629216293      - `BetaManagedAgentsSearchResultBlock object`1629416295        A block containing a web search result.1629616297        - `type: "search_result"`1629816299        - `citations: BetaManagedAgentsSearchResultCitations`1630016301          Citation settings for a search result.1630216303          - `enabled: boolean`1630416305            Whether citations are enabled for this search result.1630616307        - `content: array of BetaManagedAgentsSearchResultContent`1630816309          Array of text content blocks from the search result.1631016311          - `type: "text"`1631216313          - `text: string`1631416315            The text content.1631616317            minLength: 11631816319        - `source: string`1632016321          The URL source of the search result.1632216323          minLength: 11632416325        - `title: string`1632616327          The title of the search result.1632816329          minLength: 11633016331    - `is_error: optional boolean or null`1633216333      Whether the tool execution resulted in an error.1633416335    - `processed_at: optional string or null`1633616337      A timestamp in RFC 3339 format1633816339      format: date-time1634016341    - `session_thread_id: optional string or null`1634216343      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1634416345  - `BetaManagedAgentsAgentCustomToolUseEvent object`1634616347    Event emitted when the agent calls a custom tool. The session goes idle until the client sends a `user.custom_tool_result` event with the result.1634816349    - `type: "agent.custom_tool_use"`1635016351    - `id: string`1635216353      Unique identifier for this event.1635416355    - `input: map[unknown]`1635616357      Input parameters for the tool call.1635816359    - `name: string`1636016361      Name of the custom tool being called.1636216363    - `processed_at: string`1636416365      A timestamp in RFC 3339 format1636616367      format: date-time1636816369    - `session_thread_id: optional string or null`1637016371      When set, this event was cross-posted from a subagent's thread to surface its custom tool use on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.custom_tool_result` by `custom_tool_use_id`, so clients do not send it back.1637216373  - `BetaManagedAgentsAgentMessageEvent object`1637416375    An agent response event in the session conversation.1637616377    - `type: "agent.message"`1637816379    - `id: string`1638016381      Unique identifier for this event.1638216383    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`1638416385      Array of text blocks comprising the agent response.1638616387      - `BetaManagedAgentsTextBlock object`1638816389        Regular text content.1639016391      - `BetaManagedAgentsRedactedBlock object`1639216393        Placeholder for content withheld by Anthropic model policy.1639416395    - `processed_at: string`1639616397      A timestamp in RFC 3339 format1639816399      format: date-time1640016401  - `BetaManagedAgentsAgentThinkingEvent object`1640216403    Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.1640416405    - `type: "agent.thinking"`1640616407    - `id: string`1640816409      Unique identifier for this event.1641016411    - `processed_at: string`1641216413      A timestamp in RFC 3339 format1641416415      format: date-time1641616417  - `BetaManagedAgentsAgentMCPToolUseEvent object`1641816419    Event emitted when the agent invokes a tool provided by an MCP server.1642016421    - `type: "agent.mcp_tool_use"`1642216423    - `id: string`1642416425      Unique identifier for this event.1642616427    - `input: map[unknown]`1642816429      Input parameters for the tool call.1643016431    - `mcp_server_name: string`1643216433      Name of the MCP server providing the tool.1643416435    - `name: string`1643616437      Name of the MCP tool being used.1643816439    - `processed_at: string`1644016441      A timestamp in RFC 3339 format1644216443      format: date-time1644416445    - `evaluated_permission: optional "allow" or "ask" or "deny"`1644616447      AgentEvaluatedPermission enum1644816449      - `"allow"`1645016451      - `"ask"`1645216453      - `"deny"`1645416455    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`1645616457      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.1645816459      - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`1646016461        The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".1646216463        - `type: "always_allow"`1646416465      - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`1646616467        The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".1646816469        - `type: "always_ask"`1647016471      - `BetaManagedAgentsAgentToolEvaluationAuto object`1647216473        The resolved permission_policy was auto: the server judged this invocation individually.1647416475        - `type: "auto"`1647616477        - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`1647816479          The server's per-invocation judgement under the auto permission policy. Its type always equals the event's top-level evaluated_permission. Open union: clients must tolerate unknown variants.1648016481          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`1648216483            The server judged the invocation safe to execute without client approval.1648416485            - `type: "allow"`1648616487          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`1648816489            The server reached no judgement; the invocation is held for client approval.1649016491            - `type: "ask"`1649216493            - `reason_code: string`1649416495              The judgement's grounds in registry-bound terms, for client branching and audit rather than end-user display. Open registry; currently "indeterminate" (no judgement was reached). Clients must tolerate values outside this set.1649616497              maxLength: 641649816499          - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`1650016501            The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.1650216503            - `type: "deny"`1650416505            - `reason_code: string`1650616507              The judgement's grounds in registry-bound terms. Open registry; currently "high_risk" (judged high-risk; the call does not run). Clients must tolerate values outside this set.1650816509              maxLength: 641651016511    - `session_thread_id: optional string or null`1651216513      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` by `tool_use_id`, so clients do not send it back.1651416515  - `BetaManagedAgentsAgentMCPToolResultEvent object`1651616517    Event representing the result of an MCP tool execution.1651816519    - `type: "agent.mcp_tool_result"`1652016521    - `id: string`1652216523      Unique identifier for this event.1652416525    - `mcp_tool_use_id: string`1652616527      The id of the `agent.mcp_tool_use` event this result corresponds to.1652816529    - `processed_at: string`1653016531      A timestamp in RFC 3339 format1653216533      format: date-time1653416535    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1653616537      The result content returned by the tool.1653816539      - `BetaManagedAgentsTextBlock object`1654016541        Regular text content.1654216543      - `BetaManagedAgentsImageBlock object`1654416545        Image content specified directly as base64 data or as a reference via a URL.1654616547      - `BetaManagedAgentsDocumentBlock object`1654816549        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1655016551      - `BetaManagedAgentsSearchResultBlock object`1655216553        A block containing a web search result.1655416555    - `is_error: optional boolean or null`1655616557      Whether the tool execution resulted in an error.1655816559  - `BetaManagedAgentsAgentToolUseEvent object`1656016561    Event emitted when the agent invokes a built-in agent tool.1656216563    - `type: "agent.tool_use"`1656416565    - `id: string`1656616567      Unique identifier for this event.1656816569    - `input: map[unknown]`1657016571      Input parameters for the tool call.1657216573    - `name: string`1657416575      Name of the agent tool being used.1657616577    - `processed_at: string`1657816579      A timestamp in RFC 3339 format1658016581      format: date-time1658216583    - `evaluated_permission: optional "allow" or "ask" or "deny"`1658416585      AgentEvaluatedPermission enum1658616587      - `"allow"`1658816589      - `"ask"`1659016591      - `"deny"`1659216593    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`1659416595      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.1659616597    - `session_thread_id: optional string or null`1659816599      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` or `user.tool_result` by `tool_use_id`, so clients do not send it back.1660016601  - `BetaManagedAgentsAgentToolResultEvent object`1660216603    Event representing the result of an agent tool execution.1660416605    - `type: "agent.tool_result"`1660616607    - `id: string`1660816609      Unique identifier for this event.1661016611    - `processed_at: string`1661216613      A timestamp in RFC 3339 format1661416615      format: date-time1661616617    - `tool_use_id: string`1661816619      The id of the `agent.tool_use` event this result corresponds to.1662016621    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1662216623      The result content returned by the tool.1662416625      - `BetaManagedAgentsTextBlock object`1662616627        Regular text content.1662816629      - `BetaManagedAgentsImageBlock object`1663016631        Image content specified directly as base64 data or as a reference via a URL.1663216633      - `BetaManagedAgentsDocumentBlock object`1663416635        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1663616637      - `BetaManagedAgentsSearchResultBlock object`1663816639        A block containing a web search result.1664016641    - `is_error: optional boolean or null`1664216643      Whether the tool execution resulted in an error.1664416645  - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`1664616647    Delivery event written to the target thread's input stream when an agent-to-agent message arrives.1664816649    - `type: "agent.thread_message_received"`1665016651    - `id: string`1665216653      Unique identifier for this event.1665416655    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1665616657      Message content blocks.1665816659      - `BetaManagedAgentsTextBlock object`1666016661        Regular text content.1666216663      - `BetaManagedAgentsImageBlock object`1666416665        Image content specified directly as base64 data or as a reference via a URL.1666616667      - `BetaManagedAgentsDocumentBlock object`1666816669        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1667016671      - `BetaManagedAgentsRedactedBlock object`1667216673        Placeholder for content withheld by Anthropic model policy.1667416675    - `from_session_thread_id: string`1667616677      Public `sthr_` ID of the thread that sent the message.1667816679    - `processed_at: string`1668016681      A timestamp in RFC 3339 format1668216683      format: date-time1668416685    - `from_agent_name: optional string or null`1668616687      Name of the callable agent this message came from. Absent when received from the primary agent.1668816689  - `BetaManagedAgentsAgentThreadMessageSentEvent object`1669016691    Observability event emitted to the sender's output stream when an agent-to-agent message is sent.1669216693    - `type: "agent.thread_message_sent"`1669416695    - `id: string`1669616697      Unique identifier for this event.1669816699    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`1670016701      Message content blocks.1670216703      - `BetaManagedAgentsTextBlock object`1670416705        Regular text content.1670616707      - `BetaManagedAgentsImageBlock object`1670816709        Image content specified directly as base64 data or as a reference via a URL.1671016711      - `BetaManagedAgentsDocumentBlock object`1671216713        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1671416715      - `BetaManagedAgentsRedactedBlock object`1671616717        Placeholder for content withheld by Anthropic model policy.1671816719    - `processed_at: string`1672016721      A timestamp in RFC 3339 format1672216723      format: date-time1672416725    - `to_session_thread_id: string`1672616727      Public `sthr_` ID of the thread the message was sent to.1672816729    - `to_agent_name: optional string or null`1673016731      Name of the callable agent this message was sent to. Absent when sent to the primary agent.1673216733  - `BetaManagedAgentsAgentThreadContextCompactedEvent object`1673416735    Indicates that context compaction (summarization) occurred during the session.1673616737    - `type: "agent.thread_context_compacted"`1673816739    - `id: string`1674016741      Unique identifier for this event.1674216743    - `processed_at: string`1674416745      A timestamp in RFC 3339 format1674616747      format: date-time1674816749  - `BetaManagedAgentsSessionErrorEvent object`1675016751    An error event indicating a problem occurred during session execution.1675216753    - `type: "session.error"`1675416755    - `id: string`1675616757      Unique identifier for this event.1675816759    - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`1676016761      - `BetaManagedAgentsUnknownError object`1676216763        An unknown or unexpected error occurred during session execution. A fallback variant; clients that don't recognize a new error code can match on `retry_status` and `message` alone.1676416765        - `type: "unknown_error"`1676616767        - `message: string`1676816769          Human-readable error description.1677016771        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1677216773          What the client should do next in response to this error.1677416775          - `BetaManagedAgentsRetryStatusRetrying object`1677616777            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1677816779            - `type: "retrying"`1678016781          - `BetaManagedAgentsRetryStatusExhausted object`1678216783            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1678416785            - `type: "exhausted"`1678616787          - `BetaManagedAgentsRetryStatusTerminal object`1678816789            The session encountered a terminal error and will transition to `terminated` state.1679016791            - `type: "terminal"`1679216793      - `BetaManagedAgentsModelOverloadedError object`1679416795        The model is currently overloaded. Emitted after automatic retries are exhausted.1679616797        - `type: "model_overloaded_error"`1679816799        - `message: string`1680016801          Human-readable error description.1680216803        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1680416805          What the client should do next in response to this error.1680616807          - `BetaManagedAgentsRetryStatusRetrying object`1680816809            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1681016811          - `BetaManagedAgentsRetryStatusExhausted object`1681216813            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1681416815          - `BetaManagedAgentsRetryStatusTerminal object`1681616817            The session encountered a terminal error and will transition to `terminated` state.1681816819      - `BetaManagedAgentsModelRateLimitedError object`1682016821        The model request was rate-limited.1682216823        - `type: "model_rate_limited_error"`1682416825        - `message: string`1682616827          Human-readable error description.1682816829        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1683016831          What the client should do next in response to this error.1683216833          - `BetaManagedAgentsRetryStatusRetrying object`1683416835            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1683616837          - `BetaManagedAgentsRetryStatusExhausted object`1683816839            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1684016841          - `BetaManagedAgentsRetryStatusTerminal object`1684216843            The session encountered a terminal error and will transition to `terminated` state.1684416845      - `BetaManagedAgentsModelRequestFailedError object`1684616847        A model request failed for a reason other than overload or rate-limiting.1684816849        - `type: "model_request_failed_error"`1685016851        - `message: string`1685216853          Human-readable error description.1685416855        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1685616857          What the client should do next in response to this error.1685816859          - `BetaManagedAgentsRetryStatusRetrying object`1686016861            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1686216863          - `BetaManagedAgentsRetryStatusExhausted object`1686416865            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1686616867          - `BetaManagedAgentsRetryStatusTerminal object`1686816869            The session encountered a terminal error and will transition to `terminated` state.1687016871      - `BetaManagedAgentsMCPConnectionFailedError object`1687216873        Failed to connect to an MCP server.1687416875        - `type: "mcp_connection_failed_error"`1687616877        - `mcp_server_name: string`1687816879          Name of the MCP server that failed to connect.1688016881        - `message: string`1688216883          Human-readable error description.1688416885        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1688616887          What the client should do next in response to this error.1688816889          - `BetaManagedAgentsRetryStatusRetrying object`1689016891            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1689216893          - `BetaManagedAgentsRetryStatusExhausted object`1689416895            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1689616897          - `BetaManagedAgentsRetryStatusTerminal object`1689816899            The session encountered a terminal error and will transition to `terminated` state.1690016901      - `BetaManagedAgentsMCPAuthenticationFailedError object`1690216903        Authentication to an MCP server failed.1690416905        - `type: "mcp_authentication_failed_error"`1690616907        - `mcp_server_name: string`1690816909          Name of the MCP server that failed authentication.1691016911        - `message: string`1691216913          Human-readable error description.1691416915        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1691616917          What the client should do next in response to this error.1691816919          - `BetaManagedAgentsRetryStatusRetrying object`1692016921            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1692216923          - `BetaManagedAgentsRetryStatusExhausted object`1692416925            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1692616927          - `BetaManagedAgentsRetryStatusTerminal object`1692816929            The session encountered a terminal error and will transition to `terminated` state.1693016931      - `BetaManagedAgentsBillingError object`1693216933        The caller's organization or workspace cannot make model requests — out of credits or spend limit reached. Retrying with the same credentials will not succeed; the caller must resolve the billing state.1693416935        - `type: "billing_error"`1693616937        - `message: string`1693816939          Human-readable error description.1694016941        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1694216943          What the client should do next in response to this error.1694416945          - `BetaManagedAgentsRetryStatusRetrying object`1694616947            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1694816949          - `BetaManagedAgentsRetryStatusExhausted object`1695016951            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1695216953          - `BetaManagedAgentsRetryStatusTerminal object`1695416955            The session encountered a terminal error and will transition to `terminated` state.1695616957      - `BetaManagedAgentsCredentialHostUnreachableError object`1695816959        An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.1696016961        - `type: "credential_host_unreachable_error"`1696216963        - `credential_id: string`1696416965          ID of the affected credential.1696616967        - `message: string`1696816969          Human-readable error description.1697016971        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`1697216973          What the client should do next in response to this error.1697416975          - `BetaManagedAgentsRetryStatusRetrying object`1697616977            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.1697816979          - `BetaManagedAgentsRetryStatusExhausted object`1698016981            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.1698216983          - `BetaManagedAgentsRetryStatusTerminal object`1698416985            The session encountered a terminal error and will transition to `terminated` state.1698616987        - `vault_id: string`1698816989          ID of the vault containing the affected credential.1699016991    - `processed_at: string`1699216993      A timestamp in RFC 3339 format1699416995      format: date-time1699616997  - `BetaManagedAgentsSessionStatusRescheduledEvent object`1699816999    Indicates the session is recovering from an error state and is rescheduled for execution.1700017001    - `type: "session.status_rescheduled"`1700217003    - `id: string`1700417005      Unique identifier for this event.1700617007    - `processed_at: string`1700817009      A timestamp in RFC 3339 format1701017011      format: date-time1701217013  - `BetaManagedAgentsSessionStatusRunningEvent object`1701417015    Indicates the session is actively running and the agent is working.1701617017    - `type: "session.status_running"`1701817019    - `id: string`1702017021      Unique identifier for this event.1702217023    - `processed_at: string`1702417025      A timestamp in RFC 3339 format1702617027      format: date-time1702817029  - `BetaManagedAgentsSessionStatusIdleEvent object`1703017031    Indicates the agent has paused and is awaiting user input.1703217033    - `type: "session.status_idle"`1703417035    - `id: string`1703617037      Unique identifier for this event.1703817039    - `processed_at: string`1704017041      A timestamp in RFC 3339 format1704217043      format: date-time1704417045    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`1704617047      - `BetaManagedAgentsSessionEndTurn object`1704817049        The agent completed its turn naturally and is ready for the next user message.1705017051        - `type: "end_turn"`1705217053      - `BetaManagedAgentsSessionRequiresAction object`1705417055        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.1705617057        - `type: "requires_action"`1705817059        - `event_ids: array of string`1706017061          The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.1706217063      - `BetaManagedAgentsSessionRetriesExhausted object`1706417065        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.1706617067        - `type: "retries_exhausted"`1706817069      - `BetaManagedAgentsSessionBudgetReached object`1707017071        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.1707217073        - `type: "budget_reached"`1707417075  - `BetaManagedAgentsSessionStatusTerminatedEvent object`1707617077    Indicates the session has terminated, either due to an error or completion.1707817079    - `type: "session.status_terminated"`1708017081    - `id: string`1708217083      Unique identifier for this event.1708417085    - `processed_at: string`1708617087      A timestamp in RFC 3339 format1708817089      format: date-time1709017091  - `BetaManagedAgentsSessionThreadCreatedEvent object`1709217093    Emitted when a subagent is spawned as a new thread. Written to the parent thread's output stream so clients observing the session see child creation.1709417095    - `type: "session.thread_created"`1709617097    - `id: string`1709817099      Unique identifier for this event.1710017101    - `agent_name: string`1710217103      Name of the callable agent the thread runs.1710417105    - `processed_at: string`1710617107      A timestamp in RFC 3339 format1710817109      format: date-time1711017111    - `session_thread_id: string`1711217113      Public `sthr_` ID of the newly created thread.1711417115  - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`1711617117    Emitted when an outcome evaluation cycle begins.1711817119    - `type: "span.outcome_evaluation_start"`1712017121    - `id: string`1712217123      Unique identifier for this event.1712417125    - `iteration: number`1712617127      0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.1712817129      format: int321713017131    - `outcome_id: string`1713217133      The `outc_` ID of the outcome being evaluated.1713417135    - `processed_at: string`1713617137      A timestamp in RFC 3339 format1713817139      format: date-time1714017141  - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`1714217143    Emitted when an outcome evaluation cycle completes. Carries the verdict and aggregate token usage. A verdict of `needs_revision` means another evaluation cycle follows; `satisfied`, `max_iterations_reached`, `failed`, or `interrupted` are terminal — no further evaluation cycles follow.1714417145    - `type: "span.outcome_evaluation_end"`1714617147    - `id: string`1714817149      Unique identifier for this event.1715017151    - `explanation: string`1715217153      Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.1715417155    - `iteration: number`1715617157      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.1715817159      format: int321716017161    - `outcome_evaluation_start_id: string`1716217163      The id of the corresponding `span.outcome_evaluation_start` event.1716417165    - `outcome_id: string`1716617167      The `outc_` ID of the outcome being evaluated.1716817169    - `processed_at: string`1717017171      A timestamp in RFC 3339 format1717217173      format: date-time1717417175    - `result: string`1717617177      Evaluation verdict. 'satisfied': criteria met, session goes idle. 'needs_revision': criteria not met, another revision cycle follows. 'max_iterations_reached': evaluation budget exhausted with criteria still unmet — one final acknowledgment turn follows before the session goes idle, but no further evaluation runs. 'failed': grader determined the rubric does not apply to the deliverables. 'interrupted': user sent an interrupt while evaluation was in progress.1717817179    - `usage: BetaManagedAgentsSpanModelUsage`1718017181      Token usage for a single model request.1718217183      - `cache_creation_input_tokens: number`1718417185        Tokens used to create prompt cache in this request.1718617187        format: int321718817189      - `cache_read_input_tokens: number`1719017191        Tokens read from prompt cache in this request.1719217193        format: int321719417195      - `input_tokens: number`1719617197        Input tokens consumed by this request.1719817199        format: int321720017201      - `output_tokens: number`1720217203        Output tokens generated by this request.1720417205        format: int321720617207      - `speed: optional "standard" or "fast" or null`1720817209        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1721017211        - `"standard"`1721217213        - `"fast"`1721417215  - `BetaManagedAgentsSpanModelRequestStartEvent object`1721617217    Emitted when a model request is initiated by the agent.1721817219    - `type: "span.model_request_start"`1722017221    - `id: string`1722217223      Unique identifier for this event.1722417225    - `processed_at: string`1722617227      A timestamp in RFC 3339 format1722817229      format: date-time1723017231  - `BetaManagedAgentsSpanModelRequestEndEvent object`1723217233    Emitted when a model request completes.1723417235    - `type: "span.model_request_end"`1723617237    - `id: string`1723817239      Unique identifier for this event.1724017241    - `is_error: boolean or null`1724217243      Whether the model request resulted in an error.1724417245    - `model_request_start_id: string`1724617247      The id of the corresponding `span.model_request_start` event.1724817249    - `model_usage: BetaManagedAgentsSpanModelUsage`1725017251      Token usage for a single model request.1725217253    - `processed_at: string`1725417255      A timestamp in RFC 3339 format1725617257      format: date-time1725817259  - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`1726017261    Periodic heartbeat emitted while an outcome evaluation cycle is in progress. Distinguishes 'evaluation is actively running' from 'evaluation is stuck' between the corresponding `span.outcome_evaluation_start` and `span.outcome_evaluation_end` events.1726217263    - `type: "span.outcome_evaluation_ongoing"`1726417265    - `id: string`1726617267      Unique identifier for this event.1726817269    - `iteration: number`1727017271      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.1727217273      format: int321727417275    - `outcome_id: string`1727617277      The `outc_` ID of the outcome being evaluated.1727817279    - `processed_at: string`1728017281      A timestamp in RFC 3339 format1728217283      format: date-time1728417285  - `BetaManagedAgentsUserDefineOutcomeEvent object`1728617287    Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.1728817289    - `type: "user.define_outcome"`1729017291    - `id: string`1729217293      Unique identifier for this event.1729417295    - `description: string`1729617297      What the agent should produce. Copied from the input event.1729817299    - `max_iterations: number or null`1730017301      Evaluate-then-revise cycles before giving up. Default 3, max 20.1730217303      format: int321730417305    - `outcome_id: string`1730617307      Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.1730817309    - `processed_at: string`1731017311      A timestamp in RFC 3339 format1731217313      format: date-time1731417315    - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`1731617317      Rubric for grading the quality of an outcome.1731817319      - `BetaManagedAgentsFileRubric object`1732017321        Rubric referenced by a file uploaded via the Files API.1732217323        - `type: "file"`1732417325        - `file_id: string`1732617327          ID of the rubric file.1732817329      - `BetaManagedAgentsTextRubric object`1733017331        Rubric content provided inline as text.1733217333        - `type: "text"`1733417335        - `content: string`1733617337          Rubric content. Plain text or markdown — the grader treats it as freeform text.1733817339  - `BetaManagedAgentsSessionDeletedEvent object`1734017341    Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.1734217343    - `type: "session.deleted"`1734417345    - `id: string`1734617347      Unique identifier for this event.1734817349    - `processed_at: string`1735017351      A timestamp in RFC 3339 format1735217353      format: date-time1735417355  - `BetaManagedAgentsSessionThreadStatusRunningEvent object`1735617357    A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1735817359    - `type: "session.thread_status_running"`1736017361    - `id: string`1736217363      Unique identifier for this event.1736417365    - `agent_name: string`1736617367      Name of the agent the thread runs.1736817369    - `processed_at: string`1737017371      A timestamp in RFC 3339 format1737217373      format: date-time1737417375    - `session_thread_id: string`1737617377      Public sthr_ ID of the thread that started running.1737817379  - `BetaManagedAgentsSessionThreadStatusIdleEvent object`1738017381    A session thread has yielded and is awaiting input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1738217383    - `type: "session.thread_status_idle"`1738417385    - `id: string`1738617387      Unique identifier for this event.1738817389    - `agent_name: string`1739017391      Name of the agent the thread runs.1739217393    - `processed_at: string`1739417395      A timestamp in RFC 3339 format1739617397      format: date-time1739817399    - `session_thread_id: string`1740017401      Public sthr_ ID of the thread that went idle.1740217403    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`1740417405      - `BetaManagedAgentsSessionEndTurn object`1740617407        The agent completed its turn naturally and is ready for the next user message.1740817409      - `BetaManagedAgentsSessionRequiresAction object`1741017411        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.1741217413      - `BetaManagedAgentsSessionRetriesExhausted object`1741417415        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.1741617417      - `BetaManagedAgentsSessionBudgetReached object`1741817419        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.1742017421  - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`1742217423    A session thread has terminated and will accept no further input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1742417425    - `type: "session.thread_status_terminated"`1742617427    - `id: string`1742817429      Unique identifier for this event.1743017431    - `agent_name: string`1743217433      Name of the agent the thread runs.1743417435    - `processed_at: string`1743617437      A timestamp in RFC 3339 format1743817439      format: date-time1744017441    - `session_thread_id: string`1744217443      Public sthr_ ID of the thread that terminated.1744417445  - `BetaManagedAgentsUserToolResultEvent object`1744617447    Event sent by the client providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.1744817449    - `type: "user.tool_result"`1745017451    - `id: string`1745217453      Unique identifier for this event.1745417455    - `tool_use_id: string`1745617457      The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.1745817459    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`1746017461      The result content returned by the tool.1746217463      - `BetaManagedAgentsTextBlock object`1746417465        Regular text content.1746617467      - `BetaManagedAgentsImageBlock object`1746817469        Image content specified directly as base64 data or as a reference via a URL.1747017471      - `BetaManagedAgentsDocumentBlock object`1747217473        Document content, either specified directly as base64 data, as text, or as a reference via a URL.1747417475      - `BetaManagedAgentsSearchResultBlock object`1747617477        A block containing a web search result.1747817479    - `is_error: optional boolean or null`1748017481      Whether the tool execution resulted in an error.1748217483    - `processed_at: optional string or null`1748417485      A timestamp in RFC 3339 format1748617487      format: date-time1748817489    - `session_thread_id: optional string or null`1749017491      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.1749217493  - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`1749417495    A session thread hit a transient error and is retrying automatically. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.1749617497    - `type: "session.thread_status_rescheduled"`1749817499    - `id: string`1750017501      Unique identifier for this event.1750217503    - `agent_name: string`1750417505      Name of the agent the thread runs.1750617507    - `processed_at: string`1750817509      A timestamp in RFC 3339 format1751017511      format: date-time1751217513    - `session_thread_id: string`1751417515      Public sthr_ ID of the thread that is retrying.1751617517  - `BetaManagedAgentsSessionUpdatedEvent object`1751817519    Emitted when an UpdateSession request changed at least one field. Carries only the fields that changed; absent fields were not part of the update. The new configuration applies from the next turn.1752017521    - `type: "session.updated"`1752217523    - `id: string`1752417525      Unique identifier for this event.1752617527    - `processed_at: string`1752817529      A timestamp in RFC 3339 format1753017531      format: date-time1753217533    - `agent: optional BetaManagedAgentsSessionAgent or null`1753417535      Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.1753617537      - `type: "agent"`1753817539      - `id: string`1754017541      - `description: string or null`1754217543      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1754417545        - `type: "url"`1754617547        - `name: string`1754817549        - `url: string`1755017551      - `model: BetaManagedAgentsModelConfig`1755217553        Model identifier and configuration.1755417555        - `id: BetaManagedAgentsModel`1755617557          The model that will power your agent.1755817559          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1756017561          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`1756217563            The model that will power your agent.1756417565            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1756617567            - `"claude-fable-5-1"`1756817569              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows1757017571            - `"claude-sonnet-5"`1757217573              High-performance model for coding and agents1757417575            - `"claude-fable-5"`1757617577              Next generation of intelligence for the hardest knowledge work and coding problems1757817579            - `"claude-opus-5"`1758017581              Powerful intelligence for long-running agents and coding1758217583            - `"claude-opus-4-8"`1758417585              Powerful intelligence for long-running agents and coding1758617587            - `"claude-opus-4-7"`1758817589              Powerful intelligence for long-running agents and coding1759017591            - `"claude-opus-4-6"`1759217593              Powerful intelligence for long-running agents and coding1759417595            - `"claude-sonnet-4-6"`1759617597              Best combination of speed and intelligence1759817599            - `"claude-haiku-4-5"`1760017601              Fastest model with near-frontier intelligence1760217603            - `"claude-haiku-4-5-20251001"`1760417605              Fastest model with near-frontier intelligence1760617607            - `"claude-opus-4-5"`1760817609              Powerful intelligence for long-running agents and coding1761017611            - `"claude-opus-4-5-20251101"`1761217613              Powerful intelligence for long-running agents and coding1761417615            - `"claude-sonnet-4-5"`1761617617              High-performance model for agents and coding1761817619            - `"claude-sonnet-4-5-20250929"`1762017621              High-performance model for agents and coding1762217623          - `string`1762417625        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`1762617627          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.1762817629          - `BetaManagedAgentsEffortLow object`1763017631            Low effort. Favors latency over reasoning depth.1763217633            - `type: "low"`1763417635          - `BetaManagedAgentsEffortMedium object`1763617637            Medium effort. Balances latency and reasoning depth.1763817639            - `type: "medium"`1764017641          - `BetaManagedAgentsEffortHigh object`1764217643            High effort. Favors reasoning depth.1764417645            - `type: "high"`1764617647          - `BetaManagedAgentsEffortXhigh object`1764817649            Extra-high effort. Not all models accept this level.1765017651            - `type: "xhigh"`1765217653          - `BetaManagedAgentsEffortMax object`1765417655            Maximum effort. Favors reasoning depth over latency.1765617657            - `type: "max"`1765817659        - `inference_geo: optional string`1766017661          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.1766217663        - `speed: optional "standard" or "fast"`1766417665          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1766617667          - `"standard"`1766817669          - `"fast"`1767017671      - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`1767217673        Resolved coordinator topology with full agent definitions for each roster member.1767417675        - `type: "coordinator"`1767617677        - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`1767817679          Full `agent` definitions the coordinator may spawn as session threads.1768017681          - `BetaManagedAgentsSessionThreadAgent object`1768217683            Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.1768417685            - `type: "agent"`1768617687            - `id: string`1768817689            - `description: string or null`1769017691            - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1769217693              - `type: "url"`1769417695              - `name: string`1769617697              - `url: string`1769817699            - `model: BetaManagedAgentsModelConfig`1770017701              Model identifier and configuration.1770217703            - `name: string`1770417705            - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1770617707              - `BetaManagedAgentsAnthropicSkill object`1770817709                A resolved Anthropic-managed skill.1771017711                - `type: "anthropic"`1771217713                - `skill_id: string`1771417715                - `version: string`1771617717              - `BetaManagedAgentsCustomSkill object`1771817719                A resolved user-created custom skill.1772017721                - `type: "custom"`1772217723                - `skill_id: string`1772417725                - `version: string`1772617727            - `system: string or null`1772817729            - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1773017731              - `BetaManagedAgentsAgentToolset20260401 object`1773217733                - `type: "agent_toolset_20260401"`1773417735                - `configs: array of BetaManagedAgentsAgentToolConfig`1773617737                  - `BetaManagedAgentsBashToolConfig object`1773817739                    Configuration for the bash tool.1774017741                    - `type: "bash"`1774217743                    - `enabled: boolean`1774417745                    - `name: "bash"`1774617747                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1774817749                      Permission policy for tool execution.1775017751                      - `BetaManagedAgentsAlwaysAllowPolicy object`1775217753                        Tool calls are automatically approved without user confirmation.1775417755                        - `type: "always_allow"`1775617757                      - `BetaManagedAgentsAlwaysAskPolicy object`1775817759                        Tool calls require user confirmation before execution.1776017761                        - `type: "always_ask"`1776217763                      - `BetaManagedAgentsAutoPolicy object`1776417765                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1776617767                        - `type: "auto"`1776817769                  - `BetaManagedAgentsEditToolConfig object`1777017771                    Configuration for the edit tool.1777217773                    - `type: "edit"`1777417775                    - `enabled: boolean`1777617777                    - `name: "edit"`1777817779                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1778017781                      Permission policy for tool execution.1778217783                      - `BetaManagedAgentsAlwaysAllowPolicy object`1778417785                        Tool calls are automatically approved without user confirmation.1778617787                      - `BetaManagedAgentsAlwaysAskPolicy object`1778817789                        Tool calls require user confirmation before execution.1779017791                      - `BetaManagedAgentsAutoPolicy object`1779217793                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1779417795                  - `BetaManagedAgentsReadToolConfig object`1779617797                    Configuration for the read tool.1779817799                    - `type: "read"`1780017801                    - `enabled: boolean`1780217803                    - `name: "read"`1780417805                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1780617807                      Permission policy for tool execution.1780817809                      - `BetaManagedAgentsAlwaysAllowPolicy object`1781017811                        Tool calls are automatically approved without user confirmation.1781217813                      - `BetaManagedAgentsAlwaysAskPolicy object`1781417815                        Tool calls require user confirmation before execution.1781617817                      - `BetaManagedAgentsAutoPolicy object`1781817819                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1782017821                  - `BetaManagedAgentsWriteToolConfig object`1782217823                    Configuration for the write tool.1782417825                    - `type: "write"`1782617827                    - `enabled: boolean`1782817829                    - `name: "write"`1783017831                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1783217833                      Permission policy for tool execution.1783417835                      - `BetaManagedAgentsAlwaysAllowPolicy object`1783617837                        Tool calls are automatically approved without user confirmation.1783817839                      - `BetaManagedAgentsAlwaysAskPolicy object`1784017841                        Tool calls require user confirmation before execution.1784217843                      - `BetaManagedAgentsAutoPolicy object`1784417845                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1784617847                  - `BetaManagedAgentsGlobToolConfig object`1784817849                    Configuration for the glob tool.1785017851                    - `type: "glob"`1785217853                    - `enabled: boolean`1785417855                    - `name: "glob"`1785617857                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1785817859                      Permission policy for tool execution.1786017861                      - `BetaManagedAgentsAlwaysAllowPolicy object`1786217863                        Tool calls are automatically approved without user confirmation.1786417865                      - `BetaManagedAgentsAlwaysAskPolicy object`1786617867                        Tool calls require user confirmation before execution.1786817869                      - `BetaManagedAgentsAutoPolicy object`1787017871                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1787217873                  - `BetaManagedAgentsGrepToolConfig object`1787417875                    Configuration for the grep tool.1787617877                    - `type: "grep"`1787817879                    - `enabled: boolean`1788017881                    - `name: "grep"`1788217883                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1788417885                      Permission policy for tool execution.1788617887                      - `BetaManagedAgentsAlwaysAllowPolicy object`1788817889                        Tool calls are automatically approved without user confirmation.1789017891                      - `BetaManagedAgentsAlwaysAskPolicy object`1789217893                        Tool calls require user confirmation before execution.1789417895                      - `BetaManagedAgentsAutoPolicy object`1789617897                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1789817899                  - `BetaManagedAgentsWebFetchToolConfig object`1790017901                    Configuration for the web_fetch tool.1790217903                    - `type: "web_fetch"`1790417905                    - `enabled: boolean`1790617907                    - `name: "web_fetch"`1790817909                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1791017911                      Permission policy for tool execution.1791217913                      - `BetaManagedAgentsAlwaysAllowPolicy object`1791417915                        Tool calls are automatically approved without user confirmation.1791617917                      - `BetaManagedAgentsAlwaysAskPolicy object`1791817919                        Tool calls require user confirmation before execution.1792017921                      - `BetaManagedAgentsAutoPolicy object`1792217923                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1792417925                    - `allowed_domains: optional array of string`1792617927                    - `blocked_domains: optional array of string`1792817929                    - `max_content_tokens: optional number or null`1793017931                      format: int321793217933                  - `BetaManagedAgentsWebSearchToolConfig object`1793417935                    Configuration for the web_search tool.1793617937                    - `type: "web_search"`1793817939                    - `enabled: boolean`1794017941                    - `name: "web_search"`1794217943                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1794417945                      Permission policy for tool execution.1794617947                      - `BetaManagedAgentsAlwaysAllowPolicy object`1794817949                        Tool calls are automatically approved without user confirmation.1795017951                      - `BetaManagedAgentsAlwaysAskPolicy object`1795217953                        Tool calls require user confirmation before execution.1795417955                      - `BetaManagedAgentsAutoPolicy object`1795617957                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1795817959                    - `allowed_domains: optional array of string`1796017961                    - `blocked_domains: optional array of string`1796217963                    - `user_location: optional BetaManagedAgentsUserLocation or null`1796417965                      Approximate user location for search result localization.1796617967                      - `type: "approximate"`1796817969                        Location precision. Only "approximate" is supported.1797017971                      - `city: optional string or null`1797217973                        City name.1797417975                        minLength: 1, maxLength: 2551797617977                      - `country: optional string or null`1797817979                        Two-letter ISO 3166-1 country code, uppercase.1798017981                      - `region: optional string or null`1798217983                        Region or state name.1798417985                        minLength: 1, maxLength: 2551798617987                      - `timezone: optional string or null`1798817989                        IANA timezone identifier, e.g. "America/Los_Angeles".1799017991                        minLength: 1, maxLength: 2551799217993                - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`1799417995                  Resolved default configuration for agent tools.1799617997                  - `enabled: boolean`1799817999                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1800018001                    Permission policy for tool execution.1800218003                    - `BetaManagedAgentsAlwaysAllowPolicy object`1800418005                      Tool calls are automatically approved without user confirmation.1800618007                    - `BetaManagedAgentsAlwaysAskPolicy object`1800818009                      Tool calls require user confirmation before execution.1801018011                    - `BetaManagedAgentsAutoPolicy object`1801218013                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1801418015              - `BetaManagedAgentsMCPToolset object`1801618017                - `type: "mcp_toolset"`1801818019                - `configs: array of BetaManagedAgentsMCPToolConfig`1802018021                  - `enabled: boolean`1802218023                  - `name: string`1802418025                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1802618027                    Permission policy for tool execution.1802818029                    - `BetaManagedAgentsAlwaysAllowPolicy object`1803018031                      Tool calls are automatically approved without user confirmation.1803218033                    - `BetaManagedAgentsAlwaysAskPolicy object`1803418035                      Tool calls require user confirmation before execution.1803618037                    - `BetaManagedAgentsAutoPolicy object`1803818039                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1804018041                - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`1804218043                  Resolved default configuration for all tools from an MCP server.1804418045                  - `enabled: boolean`1804618047                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1804818049                    Permission policy for tool execution.1805018051                    - `BetaManagedAgentsAlwaysAllowPolicy object`1805218053                      Tool calls are automatically approved without user confirmation.1805418055                    - `BetaManagedAgentsAlwaysAskPolicy object`1805618057                      Tool calls require user confirmation before execution.1805818059                    - `BetaManagedAgentsAutoPolicy object`1806018061                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1806218063                - `mcp_server_name: string`1806418065              - `BetaManagedAgentsCustomTool object`1806618067                A custom tool as returned in API responses.1806818069                - `type: "custom"`1807018071                - `description: string`1807218073                - `input_schema: BetaManagedAgentsCustomToolInputSchema`1807418075                  JSON Schema for custom tool input parameters.1807618077                  - `type: "object"`1807818079                  - `properties: optional map[unknown] or null`1808018081                  - `required: optional array of string or null`1808218083                - `name: string`1808418085            - `version: number`1808618087              format: int321808818089          - `BetaManagedAgentsAdvisor object`1809018091            Platform advisor roster entry: a model the session's primary thread may consult mid-turn.1809218093            - `type: "advisor"`1809418095            - `model: string`1809618097              The advisor model id.1809818099      - `name: string`1810018101      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1810218103        - `BetaManagedAgentsAnthropicSkill object`1810418105          A resolved Anthropic-managed skill.1810618107        - `BetaManagedAgentsCustomSkill object`1810818109          A resolved user-created custom skill.1811018111      - `system: string or null`1811218113      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1811418115        - `BetaManagedAgentsAgentToolset20260401 object`1811618117        - `BetaManagedAgentsMCPToolset object`1811818119        - `BetaManagedAgentsCustomTool object`1812018121          A custom tool as returned in API responses.1812218123      - `version: number`1812418125        format: int321812618127    - `budget: optional BetaManagedAgentsBudgetLimit or null`1812818129      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1813018131      - `type: "limit"`1813218133      - `max_list_cost: BetaMonetaryAmount`1813418135        A monetary amount in a specific currency.1813618137        - `amount: string`1813818139          Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.1814018141        - `currency: BetaCurrency`1814218143          Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.1814418145    - `metadata: optional map[string]`1814618147      The session's full metadata bag after the update. Present when the update set non-empty metadata; absent when metadata was unchanged or cleared to empty.1814818149    - `title: optional string or null`1815018151      The session's new title. Present only when the update changed it.1815218153  - `BetaManagedAgentsStartEvent object`1815418155    Opens a preview of a buffered event. Carries the previewed event's type and id only. Followed by zero or more event_delta events with the same event id, normally concluded by the buffered event carrying that id. If the producing model request ends without that event (an error or interrupt mid-stream), its terminal span.model_request_end closes the preview. Only sent on stream connections that opt in via event_deltas; never appears in event history.1815618157    - `type: "event_start"`1815818159    - `event: BetaManagedAgentsStartEventPreview`1816018161      The previewed event's type and id. The event type determines which delta types the preview's event_delta events carry: agent.message events stream content_delta fragments; agent.thinking previews are start-only — no deltas follow, and the buffered agent.thinking with the same id concludes them.1816218163      - `BetaManagedAgentsAgentMessagePreview object`1816418165        - `type: "agent.message"`1816618167        - `id: string`1816818169          The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.1817018171      - `BetaManagedAgentsAgentThinkingPreview object`1817218173        - `type: "agent.thinking"`1817418175        - `id: string`1817618177          The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.1817818179  - `BetaManagedAgentsDeltaEvent object`1818018181    An incremental update to an event that is still being streamed. Deltas are best-effort and may stop early; when the buffered event with id == event_id is produced it carries the complete content. A model request that ends early (an error or interrupt) produces no buffered event — its terminal span.model_request_end closes the preview. Only sent on stream connections that opt in via event_deltas; never appears in event history.1818218183    - `type: "event_delta"`1818418185    - `delta: BetaManagedAgentsDeltaContent`1818618187      One fragment of the previewed event. The delta type is named for the previewed event's field it streams into: agent.message events stream content_delta fragments, each a partial element of the content array.1818818189      - `type: "content_delta"`1819018191      - `content: BetaManagedAgentsTextBlock`1819218193        Regular text content.1819418195      - `index: optional number`1819618197        Which entry in the previewed event's content array this fragment lands in. Insert content as that entry when the index is new; append to the existing entry otherwise.1819818199        format: uint321820018201    - `event_id: string`1820218203      The id of the event being previewed. Matches event.id on the corresponding event_start and the buffered event that reconciles the preview.1820418205  - `BetaManagedAgentsSystemMessageEvent object`1820618207    A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.1820818209    - `type: "system.message"`1821018211    - `id: string`1821218213      Unique identifier for this event.1821418215    - `content: array of BetaManagedAgentsSystemContentBlock`1821618217      System content blocks. Text-only.1821818219      - `type: "text"`1822018221      - `text: string`1822218223        The text content.1822418225        minLength: 11822618227    - `processed_at: optional string or null`1822818229      A timestamp in RFC 3339 format1823018231      format: date-time1823218233  - `BetaManagedAgentsSessionUsageEvent object`1823418235    Periodic snapshot of the session's cumulative usage and tracked list cost.1823618237    - `type: "session.usage"`1823818239    - `id: string`1824018241      Unique identifier for this event.1824218243    - `processed_at: string`1824418245      A timestamp in RFC 3339 format1824618247      format: date-time1824818249    - `usage: BetaManagedAgentsSessionUsageSnapshot`1825018251      Point-in-time snapshot of a session's cumulative usage.1825218253      - `active_seconds: optional number`1825418255        Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once. This is the duration the session's runtime cost is priced on.1825618257        format: double1825818259      - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`1826018261        Prompt-cache creation token usage broken down by cache lifetime.1826218263        - `ephemeral_1h_input_tokens: optional number`1826418265          Tokens used to create 1-hour ephemeral cache entries.1826618267          format: int321826818269        - `ephemeral_5m_input_tokens: optional number`1827018271          Tokens used to create 5-minute ephemeral cache entries.1827218273          format: int321827418275      - `cache_read_input_tokens: optional number`1827618277        Total tokens read from prompt cache.1827818279        format: int321828018281      - `input_tokens: optional number`1828218283        Total input tokens consumed across all turns.1828418285        format: int321828618287      - `list_cost: optional BetaMonetaryAmount`1828818289        A monetary amount in a specific currency.1829018291      - `output_tokens: optional number`1829218293        Total output tokens generated across all turns.1829418295        format: int321829618297      - `server_tool_use: optional BetaManagedAgentsServerToolUsage`1829818299        Cumulative count of server-executed tool invocations, broken down by tool.1830018301        - `web_fetch_requests: optional number`1830218303          Number of server-executed web fetch requests.1830418305          format: int321830618307        - `web_search_requests: optional number`1830818309          Number of server-executed web search requests.1831018311          format: int321831218313    - `budget: optional BetaManagedAgentsBudgetLimit or null`1831418315      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1831618317#### Example1831818319```bash18320curl https://api.anthropic.com/v1/sessions/$SESSION_ID/events/stream \18321    -H 'anthropic-version: 2023-06-01' \18322    -H 'anthropic-beta: managed-agents-2026-04-01' \18323    -H "X-Api-Key: $ANTHROPIC_API_KEY"18324```1832518326##### Response (200)1832718328```json18329{18330  "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",18331  "content": [18332    {18333      "text": "Where is my order #1234?",18334      "type": "text"18335    }18336  ],18337  "type": "user.message",18338  "processed_at": "2026-03-15T10:00:00Z"18339}18340```1834118342## Sessions › Resources1834318344### Add Session Resource1834518346**POST** `/v1/sessions/{session_id}/resources`1834718348Add Session Resource1834918350#### Path parameters1835118352- `session_id: string`1835318354#### Headers1835518356- `"anthropic-beta": optional array of AnthropicBeta`1835718358  Optional header to specify the beta version(s) you want to use.1835918360  - `string`1836118362  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1836318364    - `"message-batches-2024-09-24"`1836518366    - `"prompt-caching-2024-07-31"`1836718368    - `"computer-use-2024-10-22"`1836918370    - `"computer-use-2025-01-24"`1837118372    - `"pdfs-2024-09-25"`1837318374    - `"token-counting-2024-11-01"`1837518376    - `"token-efficient-tools-2025-02-19"`1837718378    - `"output-128k-2025-02-19"`1837918380    - `"files-api-2025-04-14"`1838118382    - `"mcp-client-2025-04-04"`1838318384    - `"mcp-client-2025-11-20"`1838518386    - `"dev-full-thinking-2025-05-14"`1838718388    - `"interleaved-thinking-2025-05-14"`1838918390    - `"code-execution-2025-05-22"`1839118392    - `"extended-cache-ttl-2025-04-11"`1839318394    - `"context-1m-2025-08-07"`1839518396    - `"context-management-2025-06-27"`1839718398    - `"model-context-window-exceeded-2025-08-26"`1839918400    - `"skills-2025-10-02"`1840118402    - `"fast-mode-2026-02-01"`1840318404    - `"output-300k-2026-03-24"`1840518406    - `"user-profiles-2026-03-24"`1840718408    - `"user-profiles-2026-08-18"`1840918410    - `"user-profiles-2026-09-04"`1841118412    - `"advisor-tool-2026-03-01"`1841318414    - `"managed-agents-2026-04-01"`1841518416    - `"cache-diagnosis-2026-04-07"`1841718418    - `"dreaming-2026-04-21"`1841918420    - `"thinking-token-count-2026-05-13"`1842118422    - `"server-side-fallback-2026-06-01"`1842318424    - `"server-side-fallback-2026-07-01"`1842518426    - `"fallback-credit-2026-06-01"`1842718428    - `"fallback-credit-2026-07-01"`1842918430    - `"agent-memory-2026-07-22"`1843118432    - `"mid-conversation-tool-changes-2026-07-01"`1843318434    - `"compact-2026-01-12"`1843518436    - `"computer-use-2025-11-24"`1843718438    - `"mcp-tunnels-2026-06-22"`1843918440    - `"structured-outputs-2025-11-13"`1844118442    - `"task-budgets-2026-03-13"`1844318444    - `"thinking-display-updates-2026-08-18"`1844518446    - `"ce-user-management-2026-07-13"`1844718448    - `"mid-conversation-output-config-2026-07-01"`1844918450    - `"thinking-binding-controls-2026-08-01"`1845118452    - `"mid-conversation-system-clear-at-2026-08-21"`1845318454    - `"compact-2026-09-04"`1845518456- `"anthropic-workspace-id": optional string`1845718458#### Body parameters1845918460- `file_id: string`1846118462  ID of a previously uploaded file.1846318464  minLength: 1, maxLength: 1281846518466- `type: "file"`1846718468- `mount_path: optional string or null`1846918470  Mount path in the container. Defaults to `/mnt/session/uploads/<file_id>`.1847118472  minLength: 1, maxLength: 40961847318474#### Returns1847518476- `BetaManagedAgentsFileResource object`1847718478  - `type: "file"`1847918480  - `id: string`1848118482  - `created_at: string`1848318484    A timestamp in RFC 3339 format1848518486    format: date-time1848718488  - `file_id: string`1848918490  - `mount_path: string`1849118492  - `updated_at: string`1849318494    A timestamp in RFC 3339 format1849518496    format: date-time1849718498#### Example1849918500```bash18501curl https://api.anthropic.com/v1/sessions/$SESSION_ID/resources \18502    -H 'Content-Type: application/json' \18503    -H 'anthropic-version: 2023-06-01' \18504    -H 'anthropic-beta: managed-agents-2026-04-01' \18505    -H "X-Api-Key: $ANTHROPIC_API_KEY" \18506    -d '{18507          "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",18508          "type": "file",18509          "mount_path": "/uploads/receipt.pdf"18510        }'18511```1851218513##### Response (200)1851418515```json18516{18517  "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",18518  "created_at": "2026-03-15T10:00:00Z",18519  "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",18520  "mount_path": "/uploads/receipt.pdf",18521  "type": "file",18522  "updated_at": "2026-03-15T10:00:00Z"18523}18524```1852518526### List Session Resources1852718528**GET** `/v1/sessions/{session_id}/resources`1852918530List Session Resources1853118532#### Path parameters1853318534- `session_id: string`1853518536#### Query parameters1853718538- `limit: optional number`1853918540  Maximum number of resources to return per page (max 1000). If omitted, returns all resources.1854118542  format: int321854318544- `page: optional string`1854518546  Opaque cursor from a previous response's `next_page` field.1854718548#### Headers1854918550- `"anthropic-beta": optional array of AnthropicBeta`1855118552  Optional header to specify the beta version(s) you want to use.1855318554  - `string`1855518556  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1855718558    - `"message-batches-2024-09-24"`1855918560    - `"prompt-caching-2024-07-31"`1856118562    - `"computer-use-2024-10-22"`1856318564    - `"computer-use-2025-01-24"`1856518566    - `"pdfs-2024-09-25"`1856718568    - `"token-counting-2024-11-01"`1856918570    - `"token-efficient-tools-2025-02-19"`1857118572    - `"output-128k-2025-02-19"`1857318574    - `"files-api-2025-04-14"`1857518576    - `"mcp-client-2025-04-04"`1857718578    - `"mcp-client-2025-11-20"`1857918580    - `"dev-full-thinking-2025-05-14"`1858118582    - `"interleaved-thinking-2025-05-14"`1858318584    - `"code-execution-2025-05-22"`1858518586    - `"extended-cache-ttl-2025-04-11"`1858718588    - `"context-1m-2025-08-07"`1858918590    - `"context-management-2025-06-27"`1859118592    - `"model-context-window-exceeded-2025-08-26"`1859318594    - `"skills-2025-10-02"`1859518596    - `"fast-mode-2026-02-01"`1859718598    - `"output-300k-2026-03-24"`1859918600    - `"user-profiles-2026-03-24"`1860118602    - `"user-profiles-2026-08-18"`1860318604    - `"user-profiles-2026-09-04"`1860518606    - `"advisor-tool-2026-03-01"`1860718608    - `"managed-agents-2026-04-01"`1860918610    - `"cache-diagnosis-2026-04-07"`1861118612    - `"dreaming-2026-04-21"`1861318614    - `"thinking-token-count-2026-05-13"`1861518616    - `"server-side-fallback-2026-06-01"`1861718618    - `"server-side-fallback-2026-07-01"`1861918620    - `"fallback-credit-2026-06-01"`1862118622    - `"fallback-credit-2026-07-01"`1862318624    - `"agent-memory-2026-07-22"`1862518626    - `"mid-conversation-tool-changes-2026-07-01"`1862718628    - `"compact-2026-01-12"`1862918630    - `"computer-use-2025-11-24"`1863118632    - `"mcp-tunnels-2026-06-22"`1863318634    - `"structured-outputs-2025-11-13"`1863518636    - `"task-budgets-2026-03-13"`1863718638    - `"thinking-display-updates-2026-08-18"`1863918640    - `"ce-user-management-2026-07-13"`1864118642    - `"mid-conversation-output-config-2026-07-01"`1864318644    - `"thinking-binding-controls-2026-08-01"`1864518646    - `"mid-conversation-system-clear-at-2026-08-21"`1864718648    - `"compact-2026-09-04"`1864918650- `"anthropic-workspace-id": optional string`1865118652#### Returns1865318654- `data: array of BetaManagedAgentsSessionResource`1865518656  Resources for the session, ordered by `created_at`.1865718658  - `BetaManagedAgentsGitHubRepositoryResource object`1865918660    - `type: "github_repository"`1866118662    - `id: string`1866318664    - `created_at: string`1866518666      A timestamp in RFC 3339 format1866718668      format: date-time1866918670    - `mount_path: string`1867118672    - `updated_at: string`1867318674      A timestamp in RFC 3339 format1867518676      format: date-time1867718678    - `url: string`1867918680    - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`1868118682      - `BetaManagedAgentsBranchCheckout object`1868318684        - `type: "branch"`1868518686        - `name: string`1868718688          Branch name to check out.1868918690          minLength: 1, maxLength: 2551869118692      - `BetaManagedAgentsCommitCheckout object`1869318694        - `type: "commit"`1869518696        - `sha: string`1869718698          Full commit SHA to check out.1869918700          minLength: 7, maxLength: 641870118702  - `BetaManagedAgentsFileResource object`1870318704    - `type: "file"`1870518706    - `id: string`1870718708    - `created_at: string`1870918710      A timestamp in RFC 3339 format1871118712      format: date-time1871318714    - `file_id: string`1871518716    - `mount_path: string`1871718718    - `updated_at: string`1871918720      A timestamp in RFC 3339 format1872118722      format: date-time1872318724  - `BetaManagedAgentsMemoryStoreResource object`1872518726    A memory store attached to an agent session.1872718728    - `type: "memory_store"`1872918730    - `memory_store_id: string`1873118732      The memory store ID (memstore_...). Must belong to the caller's organization and workspace.1873318734    - `access: optional "read_write" or "read_only" or null`1873518736      Access mode for an attached memory store.1873718738      - `"read_write"`1873918740      - `"read_only"`1874118742    - `description: optional string`1874318744      Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.1874518746    - `instructions: optional string or null`1874718748      Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.1874918750      maxLength: 40961875118752    - `mount_path: optional string or null`1875318754      Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.1875518756    - `name: optional string or null`1875718758      Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.1875918760- `next_page: optional string or null`1876118762  Opaque cursor for the next page. Null when no more results.1876318764#### Example1876518766```bash18767curl https://api.anthropic.com/v1/sessions/$SESSION_ID/resources \18768    -H 'anthropic-version: 2023-06-01' \18769    -H 'anthropic-beta: managed-agents-2026-04-01' \18770    -H "X-Api-Key: $ANTHROPIC_API_KEY"18771```1877218773##### Response (200)1877418775```json18776{18777  "data": [18778    {18779      "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",18780      "created_at": "2026-03-15T10:00:00Z",18781      "file_id": "file_011CNha8iCJcU1wXNR6q4V8w",18782      "mount_path": "/uploads/receipt.pdf",18783      "type": "file",18784      "updated_at": "2026-03-15T10:00:00Z"18785    },18786    {18787      "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",18788      "created_at": "2026-03-15T10:00:00Z",18789      "mount_path": "/workspace/example-repo",18790      "type": "github_repository",18791      "updated_at": "2026-03-15T10:00:00Z",18792      "url": "https://github.com/example-org/example-repo",18793      "checkout": {18794        "name": "main",18795        "type": "branch"18796      }18797    }18798  ],18799  "next_page": "page_MjAyNS0wNS0xNFQwMDowMDowMFo="18800}18801```1880218803### Get Session Resource1880418805**GET** `/v1/sessions/{session_id}/resources/{resource_id}`1880618807Get Session Resource1880818809#### Path parameters1881018811- `session_id: string`1881218813- `resource_id: string`1881418815#### Headers1881618817- `"anthropic-beta": optional array of AnthropicBeta`1881818819  Optional header to specify the beta version(s) you want to use.1882018821  - `string`1882218823  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1882418825    - `"message-batches-2024-09-24"`1882618827    - `"prompt-caching-2024-07-31"`1882818829    - `"computer-use-2024-10-22"`1883018831    - `"computer-use-2025-01-24"`1883218833    - `"pdfs-2024-09-25"`1883418835    - `"token-counting-2024-11-01"`1883618837    - `"token-efficient-tools-2025-02-19"`1883818839    - `"output-128k-2025-02-19"`1884018841    - `"files-api-2025-04-14"`1884218843    - `"mcp-client-2025-04-04"`1884418845    - `"mcp-client-2025-11-20"`1884618847    - `"dev-full-thinking-2025-05-14"`1884818849    - `"interleaved-thinking-2025-05-14"`1885018851    - `"code-execution-2025-05-22"`1885218853    - `"extended-cache-ttl-2025-04-11"`1885418855    - `"context-1m-2025-08-07"`1885618857    - `"context-management-2025-06-27"`1885818859    - `"model-context-window-exceeded-2025-08-26"`1886018861    - `"skills-2025-10-02"`1886218863    - `"fast-mode-2026-02-01"`1886418865    - `"output-300k-2026-03-24"`1886618867    - `"user-profiles-2026-03-24"`1886818869    - `"user-profiles-2026-08-18"`1887018871    - `"user-profiles-2026-09-04"`1887218873    - `"advisor-tool-2026-03-01"`1887418875    - `"managed-agents-2026-04-01"`1887618877    - `"cache-diagnosis-2026-04-07"`1887818879    - `"dreaming-2026-04-21"`1888018881    - `"thinking-token-count-2026-05-13"`1888218883    - `"server-side-fallback-2026-06-01"`1888418885    - `"server-side-fallback-2026-07-01"`1888618887    - `"fallback-credit-2026-06-01"`1888818889    - `"fallback-credit-2026-07-01"`1889018891    - `"agent-memory-2026-07-22"`1889218893    - `"mid-conversation-tool-changes-2026-07-01"`1889418895    - `"compact-2026-01-12"`1889618897    - `"computer-use-2025-11-24"`1889818899    - `"mcp-tunnels-2026-06-22"`1890018901    - `"structured-outputs-2025-11-13"`1890218903    - `"task-budgets-2026-03-13"`1890418905    - `"thinking-display-updates-2026-08-18"`1890618907    - `"ce-user-management-2026-07-13"`1890818909    - `"mid-conversation-output-config-2026-07-01"`1891018911    - `"thinking-binding-controls-2026-08-01"`1891218913    - `"mid-conversation-system-clear-at-2026-08-21"`1891418915    - `"compact-2026-09-04"`1891618917- `"anthropic-workspace-id": optional string`1891818919#### Returns1892018921- `BetaManagedAgentsGitHubRepositoryResource object`1892218923  - `type: "github_repository"`1892418925  - `id: string`1892618927  - `created_at: string`1892818929    A timestamp in RFC 3339 format1893018931    format: date-time1893218933  - `mount_path: string`1893418935  - `updated_at: string`1893618937    A timestamp in RFC 3339 format1893818939    format: date-time1894018941  - `url: string`1894218943  - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`1894418945    - `BetaManagedAgentsBranchCheckout object`1894618947      - `type: "branch"`1894818949      - `name: string`1895018951        Branch name to check out.1895218953        minLength: 1, maxLength: 2551895418955    - `BetaManagedAgentsCommitCheckout object`1895618957      - `type: "commit"`1895818959      - `sha: string`1896018961        Full commit SHA to check out.1896218963        minLength: 7, maxLength: 641896418965- `BetaManagedAgentsFileResource object`1896618967  - `type: "file"`1896818969  - `id: string`1897018971  - `created_at: string`1897218973    A timestamp in RFC 3339 format1897418975    format: date-time1897618977  - `file_id: string`1897818979  - `mount_path: string`1898018981  - `updated_at: string`1898218983    A timestamp in RFC 3339 format1898418985    format: date-time1898618987- `BetaManagedAgentsMemoryStoreResource object`1898818989  A memory store attached to an agent session.1899018991  - `type: "memory_store"`1899218993  - `memory_store_id: string`1899418995    The memory store ID (memstore_...). Must belong to the caller's organization and workspace.1899618997  - `access: optional "read_write" or "read_only" or null`1899818999    Access mode for an attached memory store.1900019001    - `"read_write"`1900219003    - `"read_only"`1900419005  - `description: optional string`1900619007    Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.1900819009  - `instructions: optional string or null`1901019011    Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.1901219013    maxLength: 40961901419015  - `mount_path: optional string or null`1901619017    Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.1901819019  - `name: optional string or null`1902019021    Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.1902219023#### Example1902419025```bash19026curl https://api.anthropic.com/v1/sessions/$SESSION_ID/resources/$RESOURCE_ID \19027    -H 'anthropic-version: 2023-06-01' \19028    -H 'anthropic-beta: managed-agents-2026-04-01' \19029    -H "X-Api-Key: $ANTHROPIC_API_KEY"19030```1903119032##### Response (200)1903319034```json19035{19036  "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",19037  "created_at": "2026-03-15T10:00:00Z",19038  "mount_path": "/workspace/example-repo",19039  "type": "github_repository",19040  "updated_at": "2026-03-15T10:00:00Z",19041  "url": "https://github.com/example-org/example-repo",19042  "checkout": {19043    "name": "main",19044    "type": "branch"19045  }19046}19047```1904819049### Update Session Resource1905019051**POST** `/v1/sessions/{session_id}/resources/{resource_id}`1905219053Update Session Resource1905419055#### Path parameters1905619057- `session_id: string`1905819059- `resource_id: string`1906019061#### Headers1906219063- `"anthropic-beta": optional array of AnthropicBeta`1906419065  Optional header to specify the beta version(s) you want to use.1906619067  - `string`1906819069  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1907019071    - `"message-batches-2024-09-24"`1907219073    - `"prompt-caching-2024-07-31"`1907419075    - `"computer-use-2024-10-22"`1907619077    - `"computer-use-2025-01-24"`1907819079    - `"pdfs-2024-09-25"`1908019081    - `"token-counting-2024-11-01"`1908219083    - `"token-efficient-tools-2025-02-19"`1908419085    - `"output-128k-2025-02-19"`1908619087    - `"files-api-2025-04-14"`1908819089    - `"mcp-client-2025-04-04"`1909019091    - `"mcp-client-2025-11-20"`1909219093    - `"dev-full-thinking-2025-05-14"`1909419095    - `"interleaved-thinking-2025-05-14"`1909619097    - `"code-execution-2025-05-22"`1909819099    - `"extended-cache-ttl-2025-04-11"`1910019101    - `"context-1m-2025-08-07"`1910219103    - `"context-management-2025-06-27"`1910419105    - `"model-context-window-exceeded-2025-08-26"`1910619107    - `"skills-2025-10-02"`1910819109    - `"fast-mode-2026-02-01"`1911019111    - `"output-300k-2026-03-24"`1911219113    - `"user-profiles-2026-03-24"`1911419115    - `"user-profiles-2026-08-18"`1911619117    - `"user-profiles-2026-09-04"`1911819119    - `"advisor-tool-2026-03-01"`1912019121    - `"managed-agents-2026-04-01"`1912219123    - `"cache-diagnosis-2026-04-07"`1912419125    - `"dreaming-2026-04-21"`1912619127    - `"thinking-token-count-2026-05-13"`1912819129    - `"server-side-fallback-2026-06-01"`1913019131    - `"server-side-fallback-2026-07-01"`1913219133    - `"fallback-credit-2026-06-01"`1913419135    - `"fallback-credit-2026-07-01"`1913619137    - `"agent-memory-2026-07-22"`1913819139    - `"mid-conversation-tool-changes-2026-07-01"`1914019141    - `"compact-2026-01-12"`1914219143    - `"computer-use-2025-11-24"`1914419145    - `"mcp-tunnels-2026-06-22"`1914619147    - `"structured-outputs-2025-11-13"`1914819149    - `"task-budgets-2026-03-13"`1915019151    - `"thinking-display-updates-2026-08-18"`1915219153    - `"ce-user-management-2026-07-13"`1915419155    - `"mid-conversation-output-config-2026-07-01"`1915619157    - `"thinking-binding-controls-2026-08-01"`1915819159    - `"mid-conversation-system-clear-at-2026-08-21"`1916019161    - `"compact-2026-09-04"`1916219163- `"anthropic-workspace-id": optional string`1916419165#### Body parameters1916619167- `authorization_token: string`1916819169  New authorization token for the resource. Currently only `github_repository` resources support token rotation.1917019171  minLength: 1, maxLength: 40961917219173#### Returns1917419175- `BetaManagedAgentsGitHubRepositoryResource object`1917619177  - `type: "github_repository"`1917819179  - `id: string`1918019181  - `created_at: string`1918219183    A timestamp in RFC 3339 format1918419185    format: date-time1918619187  - `mount_path: string`1918819189  - `updated_at: string`1919019191    A timestamp in RFC 3339 format1919219193    format: date-time1919419195  - `url: string`1919619197  - `checkout: optional BetaManagedAgentsBranchCheckout or BetaManagedAgentsCommitCheckout or null`1919819199    - `BetaManagedAgentsBranchCheckout object`1920019201      - `type: "branch"`1920219203      - `name: string`1920419205        Branch name to check out.1920619207        minLength: 1, maxLength: 2551920819209    - `BetaManagedAgentsCommitCheckout object`1921019211      - `type: "commit"`1921219213      - `sha: string`1921419215        Full commit SHA to check out.1921619217        minLength: 7, maxLength: 641921819219- `BetaManagedAgentsFileResource object`1922019221  - `type: "file"`1922219223  - `id: string`1922419225  - `created_at: string`1922619227    A timestamp in RFC 3339 format1922819229    format: date-time1923019231  - `file_id: string`1923219233  - `mount_path: string`1923419235  - `updated_at: string`1923619237    A timestamp in RFC 3339 format1923819239    format: date-time1924019241- `BetaManagedAgentsMemoryStoreResource object`1924219243  A memory store attached to an agent session.1924419245  - `type: "memory_store"`1924619247  - `memory_store_id: string`1924819249    The memory store ID (memstore_...). Must belong to the caller's organization and workspace.1925019251  - `access: optional "read_write" or "read_only" or null`1925219253    Access mode for an attached memory store.1925419255    - `"read_write"`1925619257    - `"read_only"`1925819259  - `description: optional string`1926019261    Description of the memory store, snapshotted at attach time. Rendered into the agent's system prompt. Empty string when the store has no description.1926219263  - `instructions: optional string or null`1926419265    Per-attachment guidance for the agent on how to use this store. Rendered into the memory section of the system prompt. Max 4096 chars.1926619267    maxLength: 40961926819269  - `mount_path: optional string or null`1927019271    Filesystem path where the store is mounted in the session container, e.g. /mnt/memory/user-preferences. Derived from the store's name. Output-only.1927219273  - `name: optional string or null`1927419275    Display name of the memory store, snapshotted at attach time. Later edits to the store's name do not propagate to this resource.1927619277#### Example1927819279```bash19280curl https://api.anthropic.com/v1/sessions/$SESSION_ID/resources/$RESOURCE_ID \19281    -H 'Content-Type: application/json' \19282    -H 'anthropic-version: 2023-06-01' \19283    -H 'anthropic-beta: managed-agents-2026-04-01' \19284    -H "X-Api-Key: $ANTHROPIC_API_KEY" \19285    -d '{19286          "authorization_token": "ghp_exampletoken"19287        }'19288```1928919290##### Response (200)1929119292```json19293{19294  "id": "sesrsc_011CZkZCKr6eXyl0gWMOdQiu",19295  "created_at": "2026-03-15T10:00:00Z",19296  "mount_path": "/workspace/example-repo",19297  "type": "github_repository",19298  "updated_at": "2026-03-15T10:00:00Z",19299  "url": "https://github.com/example-org/example-repo",19300  "checkout": {19301    "name": "main",19302    "type": "branch"19303  }19304}19305```1930619307### Delete Session Resource1930819309**DELETE** `/v1/sessions/{session_id}/resources/{resource_id}`1931019311Delete Session Resource1931219313#### Path parameters1931419315- `session_id: string`1931619317- `resource_id: string`1931819319#### Headers1932019321- `"anthropic-beta": optional array of AnthropicBeta`1932219323  Optional header to specify the beta version(s) you want to use.1932419325  - `string`1932619327  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1932819329    - `"message-batches-2024-09-24"`1933019331    - `"prompt-caching-2024-07-31"`1933219333    - `"computer-use-2024-10-22"`1933419335    - `"computer-use-2025-01-24"`1933619337    - `"pdfs-2024-09-25"`1933819339    - `"token-counting-2024-11-01"`1934019341    - `"token-efficient-tools-2025-02-19"`1934219343    - `"output-128k-2025-02-19"`1934419345    - `"files-api-2025-04-14"`1934619347    - `"mcp-client-2025-04-04"`1934819349    - `"mcp-client-2025-11-20"`1935019351    - `"dev-full-thinking-2025-05-14"`1935219353    - `"interleaved-thinking-2025-05-14"`1935419355    - `"code-execution-2025-05-22"`1935619357    - `"extended-cache-ttl-2025-04-11"`1935819359    - `"context-1m-2025-08-07"`1936019361    - `"context-management-2025-06-27"`1936219363    - `"model-context-window-exceeded-2025-08-26"`1936419365    - `"skills-2025-10-02"`1936619367    - `"fast-mode-2026-02-01"`1936819369    - `"output-300k-2026-03-24"`1937019371    - `"user-profiles-2026-03-24"`1937219373    - `"user-profiles-2026-08-18"`1937419375    - `"user-profiles-2026-09-04"`1937619377    - `"advisor-tool-2026-03-01"`1937819379    - `"managed-agents-2026-04-01"`1938019381    - `"cache-diagnosis-2026-04-07"`1938219383    - `"dreaming-2026-04-21"`1938419385    - `"thinking-token-count-2026-05-13"`1938619387    - `"server-side-fallback-2026-06-01"`1938819389    - `"server-side-fallback-2026-07-01"`1939019391    - `"fallback-credit-2026-06-01"`1939219393    - `"fallback-credit-2026-07-01"`1939419395    - `"agent-memory-2026-07-22"`1939619397    - `"mid-conversation-tool-changes-2026-07-01"`1939819399    - `"compact-2026-01-12"`1940019401    - `"computer-use-2025-11-24"`1940219403    - `"mcp-tunnels-2026-06-22"`1940419405    - `"structured-outputs-2025-11-13"`1940619407    - `"task-budgets-2026-03-13"`1940819409    - `"thinking-display-updates-2026-08-18"`1941019411    - `"ce-user-management-2026-07-13"`1941219413    - `"mid-conversation-output-config-2026-07-01"`1941419415    - `"thinking-binding-controls-2026-08-01"`1941619417    - `"mid-conversation-system-clear-at-2026-08-21"`1941819419    - `"compact-2026-09-04"`1942019421- `"anthropic-workspace-id": optional string`1942219423#### Returns1942419425- `BetaManagedAgentsDeleteSessionResource object`1942619427  Confirmation of resource deletion.1942819429  - `type: "session_resource_deleted"`1943019431  - `id: string`1943219433#### Example1943419435```bash19436curl https://api.anthropic.com/v1/sessions/$SESSION_ID/resources/$RESOURCE_ID \19437    -X DELETE \19438    -H 'anthropic-version: 2023-06-01' \19439    -H 'anthropic-beta: managed-agents-2026-04-01' \19440    -H "X-Api-Key: $ANTHROPIC_API_KEY"19441```1944219443##### Response (200)1944419445```json19446{19447  "id": "sesrsc_011CZkZBJq5dWxk9fVLNcPht",19448  "type": "session_resource_deleted"19449}19450```1945119452## Sessions › Threads1945319454### List Session Threads1945519456**GET** `/v1/sessions/{session_id}/threads`1945719458List Session Threads1945919460#### Path parameters1946119462- `session_id: string`1946319464#### Query parameters1946519466- `limit: optional number`1946719468  Maximum results per page. Defaults to 1000.1946919470  format: int321947119472- `page: optional string`1947319474  Opaque pagination cursor from a previous response's `next_page`. Forward-only.1947519476#### Headers1947719478- `"anthropic-beta": optional array of AnthropicBeta`1947919480  Optional header to specify the beta version(s) you want to use.1948119482  - `string`1948319484  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`1948519486    - `"message-batches-2024-09-24"`1948719488    - `"prompt-caching-2024-07-31"`1948919490    - `"computer-use-2024-10-22"`1949119492    - `"computer-use-2025-01-24"`1949319494    - `"pdfs-2024-09-25"`1949519496    - `"token-counting-2024-11-01"`1949719498    - `"token-efficient-tools-2025-02-19"`1949919500    - `"output-128k-2025-02-19"`1950119502    - `"files-api-2025-04-14"`1950319504    - `"mcp-client-2025-04-04"`1950519506    - `"mcp-client-2025-11-20"`1950719508    - `"dev-full-thinking-2025-05-14"`1950919510    - `"interleaved-thinking-2025-05-14"`1951119512    - `"code-execution-2025-05-22"`1951319514    - `"extended-cache-ttl-2025-04-11"`1951519516    - `"context-1m-2025-08-07"`1951719518    - `"context-management-2025-06-27"`1951919520    - `"model-context-window-exceeded-2025-08-26"`1952119522    - `"skills-2025-10-02"`1952319524    - `"fast-mode-2026-02-01"`1952519526    - `"output-300k-2026-03-24"`1952719528    - `"user-profiles-2026-03-24"`1952919530    - `"user-profiles-2026-08-18"`1953119532    - `"user-profiles-2026-09-04"`1953319534    - `"advisor-tool-2026-03-01"`1953519536    - `"managed-agents-2026-04-01"`1953719538    - `"cache-diagnosis-2026-04-07"`1953919540    - `"dreaming-2026-04-21"`1954119542    - `"thinking-token-count-2026-05-13"`1954319544    - `"server-side-fallback-2026-06-01"`1954519546    - `"server-side-fallback-2026-07-01"`1954719548    - `"fallback-credit-2026-06-01"`1954919550    - `"fallback-credit-2026-07-01"`1955119552    - `"agent-memory-2026-07-22"`1955319554    - `"mid-conversation-tool-changes-2026-07-01"`1955519556    - `"compact-2026-01-12"`1955719558    - `"computer-use-2025-11-24"`1955919560    - `"mcp-tunnels-2026-06-22"`1956119562    - `"structured-outputs-2025-11-13"`1956319564    - `"task-budgets-2026-03-13"`1956519566    - `"thinking-display-updates-2026-08-18"`1956719568    - `"ce-user-management-2026-07-13"`1956919570    - `"mid-conversation-output-config-2026-07-01"`1957119572    - `"thinking-binding-controls-2026-08-01"`1957319574    - `"mid-conversation-system-clear-at-2026-08-21"`1957519576    - `"compact-2026-09-04"`1957719578- `"anthropic-workspace-id": optional string`1957919580#### Returns1958119582- `data: optional array of BetaManagedAgentsSessionThread`1958319584  Threads in the session, primary first then children in spawn order.1958519586  - `type: "session_thread"`1958719588  - `id: string`1958919590    Unique identifier for this thread.1959119592  - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`1959319594    The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.1959519596    - `BetaManagedAgentsSessionThreadAgent object`1959719598      Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.1959919600      - `type: "agent"`1960119602      - `id: string`1960319604      - `description: string or null`1960519606      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`1960719608        - `type: "url"`1960919610        - `name: string`1961119612        - `url: string`1961319614      - `model: BetaManagedAgentsModelConfig`1961519616        Model identifier and configuration.1961719618        - `id: BetaManagedAgentsModel`1961919620          The model that will power your agent.1962119622          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1962319624          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`1962519626            The model that will power your agent.1962719628            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.1962919630            - `"claude-fable-5-1"`1963119632              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows1963319634            - `"claude-sonnet-5"`1963519636              High-performance model for coding and agents1963719638            - `"claude-fable-5"`1963919640              Next generation of intelligence for the hardest knowledge work and coding problems1964119642            - `"claude-opus-5"`1964319644              Powerful intelligence for long-running agents and coding1964519646            - `"claude-opus-4-8"`1964719648              Powerful intelligence for long-running agents and coding1964919650            - `"claude-opus-4-7"`1965119652              Powerful intelligence for long-running agents and coding1965319654            - `"claude-opus-4-6"`1965519656              Powerful intelligence for long-running agents and coding1965719658            - `"claude-sonnet-4-6"`1965919660              Best combination of speed and intelligence1966119662            - `"claude-haiku-4-5"`1966319664              Fastest model with near-frontier intelligence1966519666            - `"claude-haiku-4-5-20251001"`1966719668              Fastest model with near-frontier intelligence1966919670            - `"claude-opus-4-5"`1967119672              Powerful intelligence for long-running agents and coding1967319674            - `"claude-opus-4-5-20251101"`1967519676              Powerful intelligence for long-running agents and coding1967719678            - `"claude-sonnet-4-5"`1967919680              High-performance model for agents and coding1968119682            - `"claude-sonnet-4-5-20250929"`1968319684              High-performance model for agents and coding1968519686          - `string`1968719688        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`1968919690          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.1969119692          - `BetaManagedAgentsEffortLow object`1969319694            Low effort. Favors latency over reasoning depth.1969519696            - `type: "low"`1969719698          - `BetaManagedAgentsEffortMedium object`1969919700            Medium effort. Balances latency and reasoning depth.1970119702            - `type: "medium"`1970319704          - `BetaManagedAgentsEffortHigh object`1970519706            High effort. Favors reasoning depth.1970719708            - `type: "high"`1970919710          - `BetaManagedAgentsEffortXhigh object`1971119712            Extra-high effort. Not all models accept this level.1971319714            - `type: "xhigh"`1971519716          - `BetaManagedAgentsEffortMax object`1971719718            Maximum effort. Favors reasoning depth over latency.1971919720            - `type: "max"`1972119722        - `inference_geo: optional string`1972319724          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.1972519726        - `speed: optional "standard" or "fast"`1972719728          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.1972919730          - `"standard"`1973119732          - `"fast"`1973319734      - `name: string`1973519736      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1973719738        - `BetaManagedAgentsAnthropicSkill object`1973919740          A resolved Anthropic-managed skill.1974119742          - `type: "anthropic"`1974319744          - `skill_id: string`1974519746          - `version: string`1974719748        - `BetaManagedAgentsCustomSkill object`1974919750          A resolved user-created custom skill.1975119752          - `type: "custom"`1975319754          - `skill_id: string`1975519756          - `version: string`1975719758      - `system: string or null`1975919760      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1976119762        - `BetaManagedAgentsAgentToolset20260401 object`1976319764          - `type: "agent_toolset_20260401"`1976519766          - `configs: array of BetaManagedAgentsAgentToolConfig`1976719768            - `BetaManagedAgentsBashToolConfig object`1976919770              Configuration for the bash tool.1977119772              - `type: "bash"`1977319774              - `enabled: boolean`1977519776              - `name: "bash"`1977719778              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1977919780                Permission policy for tool execution.1978119782                - `BetaManagedAgentsAlwaysAllowPolicy object`1978319784                  Tool calls are automatically approved without user confirmation.1978519786                  - `type: "always_allow"`1978719788                - `BetaManagedAgentsAlwaysAskPolicy object`1978919790                  Tool calls require user confirmation before execution.1979119792                  - `type: "always_ask"`1979319794                - `BetaManagedAgentsAutoPolicy object`1979519796                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1979719798                  - `type: "auto"`1979919800            - `BetaManagedAgentsEditToolConfig object`1980119802              Configuration for the edit tool.1980319804              - `type: "edit"`1980519806              - `enabled: boolean`1980719808              - `name: "edit"`1980919810              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1981119812                Permission policy for tool execution.1981319814                - `BetaManagedAgentsAlwaysAllowPolicy object`1981519816                  Tool calls are automatically approved without user confirmation.1981719818                - `BetaManagedAgentsAlwaysAskPolicy object`1981919820                  Tool calls require user confirmation before execution.1982119822                - `BetaManagedAgentsAutoPolicy object`1982319824                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1982519826            - `BetaManagedAgentsReadToolConfig object`1982719828              Configuration for the read tool.1982919830              - `type: "read"`1983119832              - `enabled: boolean`1983319834              - `name: "read"`1983519836              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1983719838                Permission policy for tool execution.1983919840                - `BetaManagedAgentsAlwaysAllowPolicy object`1984119842                  Tool calls are automatically approved without user confirmation.1984319844                - `BetaManagedAgentsAlwaysAskPolicy object`1984519846                  Tool calls require user confirmation before execution.1984719848                - `BetaManagedAgentsAutoPolicy object`1984919850                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1985119852            - `BetaManagedAgentsWriteToolConfig object`1985319854              Configuration for the write tool.1985519856              - `type: "write"`1985719858              - `enabled: boolean`1985919860              - `name: "write"`1986119862              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1986319864                Permission policy for tool execution.1986519866                - `BetaManagedAgentsAlwaysAllowPolicy object`1986719868                  Tool calls are automatically approved without user confirmation.1986919870                - `BetaManagedAgentsAlwaysAskPolicy object`1987119872                  Tool calls require user confirmation before execution.1987319874                - `BetaManagedAgentsAutoPolicy object`1987519876                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1987719878            - `BetaManagedAgentsGlobToolConfig object`1987919880              Configuration for the glob tool.1988119882              - `type: "glob"`1988319884              - `enabled: boolean`1988519886              - `name: "glob"`1988719888              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1988919890                Permission policy for tool execution.1989119892                - `BetaManagedAgentsAlwaysAllowPolicy object`1989319894                  Tool calls are automatically approved without user confirmation.1989519896                - `BetaManagedAgentsAlwaysAskPolicy object`1989719898                  Tool calls require user confirmation before execution.1989919900                - `BetaManagedAgentsAutoPolicy object`1990119902                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1990319904            - `BetaManagedAgentsGrepToolConfig object`1990519906              Configuration for the grep tool.1990719908              - `type: "grep"`1990919910              - `enabled: boolean`1991119912              - `name: "grep"`1991319914              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1991519916                Permission policy for tool execution.1991719918                - `BetaManagedAgentsAlwaysAllowPolicy object`1991919920                  Tool calls are automatically approved without user confirmation.1992119922                - `BetaManagedAgentsAlwaysAskPolicy object`1992319924                  Tool calls require user confirmation before execution.1992519926                - `BetaManagedAgentsAutoPolicy object`1992719928                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1992919930            - `BetaManagedAgentsWebFetchToolConfig object`1993119932              Configuration for the web_fetch tool.1993319934              - `type: "web_fetch"`1993519936              - `enabled: boolean`1993719938              - `name: "web_fetch"`1993919940              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1994119942                Permission policy for tool execution.1994319944                - `BetaManagedAgentsAlwaysAllowPolicy object`1994519946                  Tool calls are automatically approved without user confirmation.1994719948                - `BetaManagedAgentsAlwaysAskPolicy object`1994919950                  Tool calls require user confirmation before execution.1995119952                - `BetaManagedAgentsAutoPolicy object`1995319954                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1995519956              - `allowed_domains: optional array of string`1995719958              - `blocked_domains: optional array of string`1995919960              - `max_content_tokens: optional number or null`1996119962                format: int321996319964            - `BetaManagedAgentsWebSearchToolConfig object`1996519966              Configuration for the web_search tool.1996719968              - `type: "web_search"`1996919970              - `enabled: boolean`1997119972              - `name: "web_search"`1997319974              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1997519976                Permission policy for tool execution.1997719978                - `BetaManagedAgentsAlwaysAllowPolicy object`1997919980                  Tool calls are automatically approved without user confirmation.1998119982                - `BetaManagedAgentsAlwaysAskPolicy object`1998319984                  Tool calls require user confirmation before execution.1998519986                - `BetaManagedAgentsAutoPolicy object`1998719988                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.1998919990              - `allowed_domains: optional array of string`1999119992              - `blocked_domains: optional array of string`1999319994              - `user_location: optional BetaManagedAgentsUserLocation or null`1999519996                Approximate user location for search result localization.1999719998                - `type: "approximate"`1999920000                  Location precision. Only "approximate" is supported.2000120002                - `city: optional string or null`2000320004                  City name.2000520006                  minLength: 1, maxLength: 2552000720008                - `country: optional string or null`2000920010                  Two-letter ISO 3166-1 country code, uppercase.2001120012                - `region: optional string or null`2001320014                  Region or state name.2001520016                  minLength: 1, maxLength: 2552001720018                - `timezone: optional string or null`2001920020                  IANA timezone identifier, e.g. "America/Los_Angeles".2002120022                  minLength: 1, maxLength: 2552002320024          - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`2002520026            Resolved default configuration for agent tools.2002720028            - `enabled: boolean`2002920030            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2003120032              Permission policy for tool execution.2003320034              - `BetaManagedAgentsAlwaysAllowPolicy object`2003520036                Tool calls are automatically approved without user confirmation.2003720038              - `BetaManagedAgentsAlwaysAskPolicy object`2003920040                Tool calls require user confirmation before execution.2004120042              - `BetaManagedAgentsAutoPolicy object`2004320044                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2004520046        - `BetaManagedAgentsMCPToolset object`2004720048          - `type: "mcp_toolset"`2004920050          - `configs: array of BetaManagedAgentsMCPToolConfig`2005120052            - `enabled: boolean`2005320054            - `name: string`2005520056            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2005720058              Permission policy for tool execution.2005920060              - `BetaManagedAgentsAlwaysAllowPolicy object`2006120062                Tool calls are automatically approved without user confirmation.2006320064              - `BetaManagedAgentsAlwaysAskPolicy object`2006520066                Tool calls require user confirmation before execution.2006720068              - `BetaManagedAgentsAutoPolicy object`2006920070                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2007120072          - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`2007320074            Resolved default configuration for all tools from an MCP server.2007520076            - `enabled: boolean`2007720078            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2007920080              Permission policy for tool execution.2008120082              - `BetaManagedAgentsAlwaysAllowPolicy object`2008320084                Tool calls are automatically approved without user confirmation.2008520086              - `BetaManagedAgentsAlwaysAskPolicy object`2008720088                Tool calls require user confirmation before execution.2008920090              - `BetaManagedAgentsAutoPolicy object`2009120092                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2009320094          - `mcp_server_name: string`2009520096        - `BetaManagedAgentsCustomTool object`2009720098          A custom tool as returned in API responses.2009920100          - `type: "custom"`2010120102          - `description: string`2010320104          - `input_schema: BetaManagedAgentsCustomToolInputSchema`2010520106            JSON Schema for custom tool input parameters.2010720108            - `type: "object"`2010920110            - `properties: optional map[unknown] or null`2011120112            - `required: optional array of string or null`2011320114          - `name: string`2011520116      - `version: number`2011720118        format: int322011920120    - `BetaManagedAgentsAdvisor object`2012120122      Platform advisor roster entry: a model the session's primary thread may consult mid-turn.2012320124      - `type: "advisor"`2012520126      - `model: string`2012720128        The advisor model id.2012920130  - `archived_at: string or null`2013120132    A timestamp in RFC 3339 format2013320134    format: date-time2013520136  - `created_at: string`2013720138    A timestamp in RFC 3339 format2013920140    format: date-time2014120142  - `parent_thread_id: string or null`2014320144    Parent thread that spawned this thread. Null for the primary thread.2014520146  - `session_id: string`2014720148    The session this thread belongs to.2014920150  - `stats: BetaManagedAgentsSessionThreadStats or null`2015120152    Timing statistics for a session thread.2015320154    - `active_seconds: optional number`2015520156      Cumulative time in seconds the thread spent actively running. Excludes idle time.2015720158      format: double2015920160    - `duration_seconds: optional number`2016120162      Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.2016320164      format: double2016520166    - `startup_seconds: optional number`2016720168      Time in seconds for the thread to begin running. Zero for child threads, which start immediately.2016920170      format: double2017120172  - `status: BetaManagedAgentsSessionThreadStatus`2017320174    SessionThreadStatus enum2017520176    - `"running"`2017720178    - `"idle"`2017920180    - `"rescheduling"`2018120182    - `"terminated"`2018320184  - `updated_at: string`2018520186    A timestamp in RFC 3339 format2018720188    format: date-time2018920190  - `usage: BetaManagedAgentsSessionThreadUsage or null`2019120192    Cumulative token usage for a session thread across all turns.2019320194    - `active_seconds: optional number`2019520196      Cumulative time in seconds this thread spent in running status. Equal to `stats.active_seconds`; surfaced here so a thread's usage carries every quantity its cost is priced on.2019720198      format: double2019920200    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`2020120202      Prompt-cache creation token usage broken down by cache lifetime.2020320204      - `ephemeral_1h_input_tokens: optional number`2020520206        Tokens used to create 1-hour ephemeral cache entries.2020720208        format: int322020920210      - `ephemeral_5m_input_tokens: optional number`2021120212        Tokens used to create 5-minute ephemeral cache entries.2021320214        format: int322021520216    - `cache_read_input_tokens: optional number`2021720218      Total tokens read from prompt cache.2021920220      format: int322022120222    - `input_tokens: optional number`2022320224      Total input tokens consumed across all turns.2022520226      format: int322022720228    - `list_cost: optional BetaMonetaryAmount or null`2022920230      A monetary amount in a specific currency.2023120232      - `amount: string`2023320234        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.2023520236      - `currency: BetaCurrency`2023720238        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.2023920240    - `output_tokens: optional number`2024120242      Total output tokens generated across all turns.2024320244      format: int322024520246    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`2024720248      Cumulative count of server-executed tool invocations, broken down by tool.2024920250      - `web_fetch_requests: optional number`2025120252        Number of server-executed web fetch requests.2025320254        format: int322025520256      - `web_search_requests: optional number`2025720258        Number of server-executed web search requests.2025920260        format: int322026120262- `next_page: optional string or null`2026320264  Opaque cursor for the next page. Null when no more results.2026520266#### Example2026720268```bash20269curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads \20270    -H 'anthropic-version: 2023-06-01' \20271    -H 'anthropic-beta: managed-agents-2026-04-01' \20272    -H "X-Api-Key: $ANTHROPIC_API_KEY"20273```2027420275##### Response (200)2027620277```json20278{20279  "data": [20280    {20281      "id": "sthr_011CZkZVWa6oIjw0rgXZpnBt",20282      "agent": {20283        "id": "agent_011CZkYqphY8vELVzwCUpqiQ",20284        "description": "A focused research subagent.",20285        "mcp_servers": [20286          {20287            "name": "example-mcp",20288            "type": "url",20289            "url": "https://example-server.modelcontextprotocol.io/sse"20290          }20291        ],20292        "model": {20293          "id": "claude-opus-5",20294          "effort": {20295            "type": "low"20296          },20297          "inference_geo": "inference_geo",20298          "speed": "standard"20299        },20300        "name": "Researcher",20301        "skills": [20302          {20303            "skill_id": "xlsx",20304            "type": "anthropic",20305            "version": "1"20306          }20307        ],20308        "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",20309        "tools": [20310          {20311            "configs": [20312              {20313                "enabled": true,20314                "name": "bash",20315                "permission_policy": {20316                  "type": "always_allow"20317                },20318                "type": "bash"20319              }20320            ],20321            "default_config": {20322              "enabled": true,20323              "permission_policy": {20324                "type": "always_ask"20325              }20326            },20327            "type": "agent_toolset_20260401"20328          }20329        ],20330        "type": "agent",20331        "version": 120332      },20333      "archived_at": null,20334      "created_at": "2026-03-15T10:00:00Z",20335      "parent_thread_id": null,20336      "session_id": "sesn_011CZkZAtmR3yMPDzynEDxu7",20337      "stats": {20338        "active_seconds": 0,20339        "duration_seconds": 0,20340        "startup_seconds": 020341      },20342      "status": "idle",20343      "type": "session_thread",20344      "updated_at": "2026-03-15T10:00:00Z",20345      "usage": {20346        "active_seconds": 0,20347        "cache_creation": {20348          "ephemeral_1h_input_tokens": 0,20349          "ephemeral_5m_input_tokens": 020350        },20351        "cache_read_input_tokens": 0,20352        "input_tokens": 0,20353        "list_cost": {20354          "amount": "2500",20355          "currency": "USD"20356        },20357        "output_tokens": 0,20358        "server_tool_use": {20359          "web_fetch_requests": 0,20360          "web_search_requests": 320361        }20362      }20363    }20364  ],20365  "next_page": "page_MjAyNS0wNS0xNFQwMDowMDowMFo="20366}20367```2036820369### Get Session Thread2037020371**GET** `/v1/sessions/{session_id}/threads/{thread_id}`2037220373Get Session Thread2037420375#### Path parameters2037620377- `session_id: string`2037820379- `thread_id: string`2038020381#### Headers2038220383- `"anthropic-beta": optional array of AnthropicBeta`2038420385  Optional header to specify the beta version(s) you want to use.2038620387  - `string`2038820389  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`2039020391    - `"message-batches-2024-09-24"`2039220393    - `"prompt-caching-2024-07-31"`2039420395    - `"computer-use-2024-10-22"`2039620397    - `"computer-use-2025-01-24"`2039820399    - `"pdfs-2024-09-25"`2040020401    - `"token-counting-2024-11-01"`2040220403    - `"token-efficient-tools-2025-02-19"`2040420405    - `"output-128k-2025-02-19"`2040620407    - `"files-api-2025-04-14"`2040820409    - `"mcp-client-2025-04-04"`2041020411    - `"mcp-client-2025-11-20"`2041220413    - `"dev-full-thinking-2025-05-14"`2041420415    - `"interleaved-thinking-2025-05-14"`2041620417    - `"code-execution-2025-05-22"`2041820419    - `"extended-cache-ttl-2025-04-11"`2042020421    - `"context-1m-2025-08-07"`2042220423    - `"context-management-2025-06-27"`2042420425    - `"model-context-window-exceeded-2025-08-26"`2042620427    - `"skills-2025-10-02"`2042820429    - `"fast-mode-2026-02-01"`2043020431    - `"output-300k-2026-03-24"`2043220433    - `"user-profiles-2026-03-24"`2043420435    - `"user-profiles-2026-08-18"`2043620437    - `"user-profiles-2026-09-04"`2043820439    - `"advisor-tool-2026-03-01"`2044020441    - `"managed-agents-2026-04-01"`2044220443    - `"cache-diagnosis-2026-04-07"`2044420445    - `"dreaming-2026-04-21"`2044620447    - `"thinking-token-count-2026-05-13"`2044820449    - `"server-side-fallback-2026-06-01"`2045020451    - `"server-side-fallback-2026-07-01"`2045220453    - `"fallback-credit-2026-06-01"`2045420455    - `"fallback-credit-2026-07-01"`2045620457    - `"agent-memory-2026-07-22"`2045820459    - `"mid-conversation-tool-changes-2026-07-01"`2046020461    - `"compact-2026-01-12"`2046220463    - `"computer-use-2025-11-24"`2046420465    - `"mcp-tunnels-2026-06-22"`2046620467    - `"structured-outputs-2025-11-13"`2046820469    - `"task-budgets-2026-03-13"`2047020471    - `"thinking-display-updates-2026-08-18"`2047220473    - `"ce-user-management-2026-07-13"`2047420475    - `"mid-conversation-output-config-2026-07-01"`2047620477    - `"thinking-binding-controls-2026-08-01"`2047820479    - `"mid-conversation-system-clear-at-2026-08-21"`2048020481    - `"compact-2026-09-04"`2048220483- `"anthropic-workspace-id": optional string`2048420485#### Returns2048620487- `BetaManagedAgentsSessionThread object`2048820489  An execution thread within a `session`. Each session has one primary thread plus zero or more child threads spawned by the coordinator.2049020491  - `type: "session_thread"`2049220493  - `id: string`2049420495    Unique identifier for this thread.2049620497  - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`2049820499    The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.2050020501    - `BetaManagedAgentsSessionThreadAgent object`2050220503      Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.2050420505      - `type: "agent"`2050620507      - `id: string`2050820509      - `description: string or null`2051020511      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`2051220513        - `type: "url"`2051420515        - `name: string`2051620517        - `url: string`2051820519      - `model: BetaManagedAgentsModelConfig`2052020521        Model identifier and configuration.2052220523        - `id: BetaManagedAgentsModel`2052420525          The model that will power your agent.2052620527          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2052820529          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`2053020531            The model that will power your agent.2053220533            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2053420535            - `"claude-fable-5-1"`2053620537              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows2053820539            - `"claude-sonnet-5"`2054020541              High-performance model for coding and agents2054220543            - `"claude-fable-5"`2054420545              Next generation of intelligence for the hardest knowledge work and coding problems2054620547            - `"claude-opus-5"`2054820549              Powerful intelligence for long-running agents and coding2055020551            - `"claude-opus-4-8"`2055220553              Powerful intelligence for long-running agents and coding2055420555            - `"claude-opus-4-7"`2055620557              Powerful intelligence for long-running agents and coding2055820559            - `"claude-opus-4-6"`2056020561              Powerful intelligence for long-running agents and coding2056220563            - `"claude-sonnet-4-6"`2056420565              Best combination of speed and intelligence2056620567            - `"claude-haiku-4-5"`2056820569              Fastest model with near-frontier intelligence2057020571            - `"claude-haiku-4-5-20251001"`2057220573              Fastest model with near-frontier intelligence2057420575            - `"claude-opus-4-5"`2057620577              Powerful intelligence for long-running agents and coding2057820579            - `"claude-opus-4-5-20251101"`2058020581              Powerful intelligence for long-running agents and coding2058220583            - `"claude-sonnet-4-5"`2058420585              High-performance model for agents and coding2058620587            - `"claude-sonnet-4-5-20250929"`2058820589              High-performance model for agents and coding2059020591          - `string`2059220593        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`2059420595          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.2059620597          - `BetaManagedAgentsEffortLow object`2059820599            Low effort. Favors latency over reasoning depth.2060020601            - `type: "low"`2060220603          - `BetaManagedAgentsEffortMedium object`2060420605            Medium effort. Balances latency and reasoning depth.2060620607            - `type: "medium"`2060820609          - `BetaManagedAgentsEffortHigh object`2061020611            High effort. Favors reasoning depth.2061220613            - `type: "high"`2061420615          - `BetaManagedAgentsEffortXhigh object`2061620617            Extra-high effort. Not all models accept this level.2061820619            - `type: "xhigh"`2062020621          - `BetaManagedAgentsEffortMax object`2062220623            Maximum effort. Favors reasoning depth over latency.2062420625            - `type: "max"`2062620627        - `inference_geo: optional string`2062820629          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.2063020631        - `speed: optional "standard" or "fast"`2063220633          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.2063420635          - `"standard"`2063620637          - `"fast"`2063820639      - `name: string`2064020641      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`2064220643        - `BetaManagedAgentsAnthropicSkill object`2064420645          A resolved Anthropic-managed skill.2064620647          - `type: "anthropic"`2064820649          - `skill_id: string`2065020651          - `version: string`2065220653        - `BetaManagedAgentsCustomSkill object`2065420655          A resolved user-created custom skill.2065620657          - `type: "custom"`2065820659          - `skill_id: string`2066020661          - `version: string`2066220663      - `system: string or null`2066420665      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`2066620667        - `BetaManagedAgentsAgentToolset20260401 object`2066820669          - `type: "agent_toolset_20260401"`2067020671          - `configs: array of BetaManagedAgentsAgentToolConfig`2067220673            - `BetaManagedAgentsBashToolConfig object`2067420675              Configuration for the bash tool.2067620677              - `type: "bash"`2067820679              - `enabled: boolean`2068020681              - `name: "bash"`2068220683              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2068420685                Permission policy for tool execution.2068620687                - `BetaManagedAgentsAlwaysAllowPolicy object`2068820689                  Tool calls are automatically approved without user confirmation.2069020691                  - `type: "always_allow"`2069220693                - `BetaManagedAgentsAlwaysAskPolicy object`2069420695                  Tool calls require user confirmation before execution.2069620697                  - `type: "always_ask"`2069820699                - `BetaManagedAgentsAutoPolicy object`2070020701                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2070220703                  - `type: "auto"`2070420705            - `BetaManagedAgentsEditToolConfig object`2070620707              Configuration for the edit tool.2070820709              - `type: "edit"`2071020711              - `enabled: boolean`2071220713              - `name: "edit"`2071420715              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2071620717                Permission policy for tool execution.2071820719                - `BetaManagedAgentsAlwaysAllowPolicy object`2072020721                  Tool calls are automatically approved without user confirmation.2072220723                - `BetaManagedAgentsAlwaysAskPolicy object`2072420725                  Tool calls require user confirmation before execution.2072620727                - `BetaManagedAgentsAutoPolicy object`2072820729                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2073020731            - `BetaManagedAgentsReadToolConfig object`2073220733              Configuration for the read tool.2073420735              - `type: "read"`2073620737              - `enabled: boolean`2073820739              - `name: "read"`2074020741              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2074220743                Permission policy for tool execution.2074420745                - `BetaManagedAgentsAlwaysAllowPolicy object`2074620747                  Tool calls are automatically approved without user confirmation.2074820749                - `BetaManagedAgentsAlwaysAskPolicy object`2075020751                  Tool calls require user confirmation before execution.2075220753                - `BetaManagedAgentsAutoPolicy object`2075420755                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2075620757            - `BetaManagedAgentsWriteToolConfig object`2075820759              Configuration for the write tool.2076020761              - `type: "write"`2076220763              - `enabled: boolean`2076420765              - `name: "write"`2076620767              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2076820769                Permission policy for tool execution.2077020771                - `BetaManagedAgentsAlwaysAllowPolicy object`2077220773                  Tool calls are automatically approved without user confirmation.2077420775                - `BetaManagedAgentsAlwaysAskPolicy object`2077620777                  Tool calls require user confirmation before execution.2077820779                - `BetaManagedAgentsAutoPolicy object`2078020781                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2078220783            - `BetaManagedAgentsGlobToolConfig object`2078420785              Configuration for the glob tool.2078620787              - `type: "glob"`2078820789              - `enabled: boolean`2079020791              - `name: "glob"`2079220793              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2079420795                Permission policy for tool execution.2079620797                - `BetaManagedAgentsAlwaysAllowPolicy object`2079820799                  Tool calls are automatically approved without user confirmation.2080020801                - `BetaManagedAgentsAlwaysAskPolicy object`2080220803                  Tool calls require user confirmation before execution.2080420805                - `BetaManagedAgentsAutoPolicy object`2080620807                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2080820809            - `BetaManagedAgentsGrepToolConfig object`2081020811              Configuration for the grep tool.2081220813              - `type: "grep"`2081420815              - `enabled: boolean`2081620817              - `name: "grep"`2081820819              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2082020821                Permission policy for tool execution.2082220823                - `BetaManagedAgentsAlwaysAllowPolicy object`2082420825                  Tool calls are automatically approved without user confirmation.2082620827                - `BetaManagedAgentsAlwaysAskPolicy object`2082820829                  Tool calls require user confirmation before execution.2083020831                - `BetaManagedAgentsAutoPolicy object`2083220833                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2083420835            - `BetaManagedAgentsWebFetchToolConfig object`2083620837              Configuration for the web_fetch tool.2083820839              - `type: "web_fetch"`2084020841              - `enabled: boolean`2084220843              - `name: "web_fetch"`2084420845              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2084620847                Permission policy for tool execution.2084820849                - `BetaManagedAgentsAlwaysAllowPolicy object`2085020851                  Tool calls are automatically approved without user confirmation.2085220853                - `BetaManagedAgentsAlwaysAskPolicy object`2085420855                  Tool calls require user confirmation before execution.2085620857                - `BetaManagedAgentsAutoPolicy object`2085820859                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2086020861              - `allowed_domains: optional array of string`2086220863              - `blocked_domains: optional array of string`2086420865              - `max_content_tokens: optional number or null`2086620867                format: int322086820869            - `BetaManagedAgentsWebSearchToolConfig object`2087020871              Configuration for the web_search tool.2087220873              - `type: "web_search"`2087420875              - `enabled: boolean`2087620877              - `name: "web_search"`2087820879              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2088020881                Permission policy for tool execution.2088220883                - `BetaManagedAgentsAlwaysAllowPolicy object`2088420885                  Tool calls are automatically approved without user confirmation.2088620887                - `BetaManagedAgentsAlwaysAskPolicy object`2088820889                  Tool calls require user confirmation before execution.2089020891                - `BetaManagedAgentsAutoPolicy object`2089220893                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2089420895              - `allowed_domains: optional array of string`2089620897              - `blocked_domains: optional array of string`2089820899              - `user_location: optional BetaManagedAgentsUserLocation or null`2090020901                Approximate user location for search result localization.2090220903                - `type: "approximate"`2090420905                  Location precision. Only "approximate" is supported.2090620907                - `city: optional string or null`2090820909                  City name.2091020911                  minLength: 1, maxLength: 2552091220913                - `country: optional string or null`2091420915                  Two-letter ISO 3166-1 country code, uppercase.2091620917                - `region: optional string or null`2091820919                  Region or state name.2092020921                  minLength: 1, maxLength: 2552092220923                - `timezone: optional string or null`2092420925                  IANA timezone identifier, e.g. "America/Los_Angeles".2092620927                  minLength: 1, maxLength: 2552092820929          - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`2093020931            Resolved default configuration for agent tools.2093220933            - `enabled: boolean`2093420935            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2093620937              Permission policy for tool execution.2093820939              - `BetaManagedAgentsAlwaysAllowPolicy object`2094020941                Tool calls are automatically approved without user confirmation.2094220943              - `BetaManagedAgentsAlwaysAskPolicy object`2094420945                Tool calls require user confirmation before execution.2094620947              - `BetaManagedAgentsAutoPolicy object`2094820949                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2095020951        - `BetaManagedAgentsMCPToolset object`2095220953          - `type: "mcp_toolset"`2095420955          - `configs: array of BetaManagedAgentsMCPToolConfig`2095620957            - `enabled: boolean`2095820959            - `name: string`2096020961            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2096220963              Permission policy for tool execution.2096420965              - `BetaManagedAgentsAlwaysAllowPolicy object`2096620967                Tool calls are automatically approved without user confirmation.2096820969              - `BetaManagedAgentsAlwaysAskPolicy object`2097020971                Tool calls require user confirmation before execution.2097220973              - `BetaManagedAgentsAutoPolicy object`2097420975                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2097620977          - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`2097820979            Resolved default configuration for all tools from an MCP server.2098020981            - `enabled: boolean`2098220983            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2098420985              Permission policy for tool execution.2098620987              - `BetaManagedAgentsAlwaysAllowPolicy object`2098820989                Tool calls are automatically approved without user confirmation.2099020991              - `BetaManagedAgentsAlwaysAskPolicy object`2099220993                Tool calls require user confirmation before execution.2099420995              - `BetaManagedAgentsAutoPolicy object`2099620997                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2099820999          - `mcp_server_name: string`2100021001        - `BetaManagedAgentsCustomTool object`2100221003          A custom tool as returned in API responses.2100421005          - `type: "custom"`2100621007          - `description: string`2100821009          - `input_schema: BetaManagedAgentsCustomToolInputSchema`2101021011            JSON Schema for custom tool input parameters.2101221013            - `type: "object"`2101421015            - `properties: optional map[unknown] or null`2101621017            - `required: optional array of string or null`2101821019          - `name: string`2102021021      - `version: number`2102221023        format: int322102421025    - `BetaManagedAgentsAdvisor object`2102621027      Platform advisor roster entry: a model the session's primary thread may consult mid-turn.2102821029      - `type: "advisor"`2103021031      - `model: string`2103221033        The advisor model id.2103421035  - `archived_at: string or null`2103621037    A timestamp in RFC 3339 format2103821039    format: date-time2104021041  - `created_at: string`2104221043    A timestamp in RFC 3339 format2104421045    format: date-time2104621047  - `parent_thread_id: string or null`2104821049    Parent thread that spawned this thread. Null for the primary thread.2105021051  - `session_id: string`2105221053    The session this thread belongs to.2105421055  - `stats: BetaManagedAgentsSessionThreadStats or null`2105621057    Timing statistics for a session thread.2105821059    - `active_seconds: optional number`2106021061      Cumulative time in seconds the thread spent actively running. Excludes idle time.2106221063      format: double2106421065    - `duration_seconds: optional number`2106621067      Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.2106821069      format: double2107021071    - `startup_seconds: optional number`2107221073      Time in seconds for the thread to begin running. Zero for child threads, which start immediately.2107421075      format: double2107621077  - `status: BetaManagedAgentsSessionThreadStatus`2107821079    SessionThreadStatus enum2108021081    - `"running"`2108221083    - `"idle"`2108421085    - `"rescheduling"`2108621087    - `"terminated"`2108821089  - `updated_at: string`2109021091    A timestamp in RFC 3339 format2109221093    format: date-time2109421095  - `usage: BetaManagedAgentsSessionThreadUsage or null`2109621097    Cumulative token usage for a session thread across all turns.2109821099    - `active_seconds: optional number`2110021101      Cumulative time in seconds this thread spent in running status. Equal to `stats.active_seconds`; surfaced here so a thread's usage carries every quantity its cost is priced on.2110221103      format: double2110421105    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`2110621107      Prompt-cache creation token usage broken down by cache lifetime.2110821109      - `ephemeral_1h_input_tokens: optional number`2111021111        Tokens used to create 1-hour ephemeral cache entries.2111221113        format: int322111421115      - `ephemeral_5m_input_tokens: optional number`2111621117        Tokens used to create 5-minute ephemeral cache entries.2111821119        format: int322112021121    - `cache_read_input_tokens: optional number`2112221123      Total tokens read from prompt cache.2112421125      format: int322112621127    - `input_tokens: optional number`2112821129      Total input tokens consumed across all turns.2113021131      format: int322113221133    - `list_cost: optional BetaMonetaryAmount or null`2113421135      A monetary amount in a specific currency.2113621137      - `amount: string`2113821139        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.2114021141      - `currency: BetaCurrency`2114221143        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.2114421145    - `output_tokens: optional number`2114621147      Total output tokens generated across all turns.2114821149      format: int322115021151    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`2115221153      Cumulative count of server-executed tool invocations, broken down by tool.2115421155      - `web_fetch_requests: optional number`2115621157        Number of server-executed web fetch requests.2115821159        format: int322116021161      - `web_search_requests: optional number`2116221163        Number of server-executed web search requests.2116421165        format: int322116621167#### Example2116821169```bash21170curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID \21171    -H 'anthropic-version: 2023-06-01' \21172    -H 'anthropic-beta: managed-agents-2026-04-01' \21173    -H "X-Api-Key: $ANTHROPIC_API_KEY"21174```2117521176##### Response (200)2117721178```json21179{21180  "id": "sthr_011CZkZVWa6oIjw0rgXZpnBt",21181  "agent": {21182    "id": "agent_011CZkYqphY8vELVzwCUpqiQ",21183    "description": "A focused research subagent.",21184    "mcp_servers": [21185      {21186        "name": "example-mcp",21187        "type": "url",21188        "url": "https://example-server.modelcontextprotocol.io/sse"21189      }21190    ],21191    "model": {21192      "id": "claude-opus-5",21193      "effort": {21194        "type": "low"21195      },21196      "inference_geo": "inference_geo",21197      "speed": "standard"21198    },21199    "name": "Researcher",21200    "skills": [21201      {21202        "skill_id": "xlsx",21203        "type": "anthropic",21204        "version": "1"21205      }21206    ],21207    "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",21208    "tools": [21209      {21210        "configs": [21211          {21212            "enabled": true,21213            "name": "bash",21214            "permission_policy": {21215              "type": "always_allow"21216            },21217            "type": "bash"21218          }21219        ],21220        "default_config": {21221          "enabled": true,21222          "permission_policy": {21223            "type": "always_ask"21224          }21225        },21226        "type": "agent_toolset_20260401"21227      }21228    ],21229    "type": "agent",21230    "version": 121231  },21232  "archived_at": null,21233  "created_at": "2026-03-15T10:00:00Z",21234  "parent_thread_id": null,21235  "session_id": "sesn_011CZkZAtmR3yMPDzynEDxu7",21236  "stats": {21237    "active_seconds": 0,21238    "duration_seconds": 0,21239    "startup_seconds": 021240  },21241  "status": "idle",21242  "type": "session_thread",21243  "updated_at": "2026-03-15T10:00:00Z",21244  "usage": {21245    "active_seconds": 0,21246    "cache_creation": {21247      "ephemeral_1h_input_tokens": 0,21248      "ephemeral_5m_input_tokens": 021249    },21250    "cache_read_input_tokens": 0,21251    "input_tokens": 0,21252    "list_cost": {21253      "amount": "2500",21254      "currency": "USD"21255    },21256    "output_tokens": 0,21257    "server_tool_use": {21258      "web_fetch_requests": 0,21259      "web_search_requests": 321260    }21261  }21262}21263```2126421265### Archive Session Thread2126621267**POST** `/v1/sessions/{session_id}/threads/{thread_id}/archive`2126821269Archive Session Thread2127021271#### Path parameters2127221273- `session_id: string`2127421275- `thread_id: string`2127621277#### Headers2127821279- `"anthropic-beta": optional array of AnthropicBeta`2128021281  Optional header to specify the beta version(s) you want to use.2128221283  - `string`2128421285  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`2128621287    - `"message-batches-2024-09-24"`2128821289    - `"prompt-caching-2024-07-31"`2129021291    - `"computer-use-2024-10-22"`2129221293    - `"computer-use-2025-01-24"`2129421295    - `"pdfs-2024-09-25"`2129621297    - `"token-counting-2024-11-01"`2129821299    - `"token-efficient-tools-2025-02-19"`2130021301    - `"output-128k-2025-02-19"`2130221303    - `"files-api-2025-04-14"`2130421305    - `"mcp-client-2025-04-04"`2130621307    - `"mcp-client-2025-11-20"`2130821309    - `"dev-full-thinking-2025-05-14"`2131021311    - `"interleaved-thinking-2025-05-14"`2131221313    - `"code-execution-2025-05-22"`2131421315    - `"extended-cache-ttl-2025-04-11"`2131621317    - `"context-1m-2025-08-07"`2131821319    - `"context-management-2025-06-27"`2132021321    - `"model-context-window-exceeded-2025-08-26"`2132221323    - `"skills-2025-10-02"`2132421325    - `"fast-mode-2026-02-01"`2132621327    - `"output-300k-2026-03-24"`2132821329    - `"user-profiles-2026-03-24"`2133021331    - `"user-profiles-2026-08-18"`2133221333    - `"user-profiles-2026-09-04"`2133421335    - `"advisor-tool-2026-03-01"`2133621337    - `"managed-agents-2026-04-01"`2133821339    - `"cache-diagnosis-2026-04-07"`2134021341    - `"dreaming-2026-04-21"`2134221343    - `"thinking-token-count-2026-05-13"`2134421345    - `"server-side-fallback-2026-06-01"`2134621347    - `"server-side-fallback-2026-07-01"`2134821349    - `"fallback-credit-2026-06-01"`2135021351    - `"fallback-credit-2026-07-01"`2135221353    - `"agent-memory-2026-07-22"`2135421355    - `"mid-conversation-tool-changes-2026-07-01"`2135621357    - `"compact-2026-01-12"`2135821359    - `"computer-use-2025-11-24"`2136021361    - `"mcp-tunnels-2026-06-22"`2136221363    - `"structured-outputs-2025-11-13"`2136421365    - `"task-budgets-2026-03-13"`2136621367    - `"thinking-display-updates-2026-08-18"`2136821369    - `"ce-user-management-2026-07-13"`2137021371    - `"mid-conversation-output-config-2026-07-01"`2137221373    - `"thinking-binding-controls-2026-08-01"`2137421375    - `"mid-conversation-system-clear-at-2026-08-21"`2137621377    - `"compact-2026-09-04"`2137821379- `"anthropic-workspace-id": optional string`2138021381#### Returns2138221383- `BetaManagedAgentsSessionThread object`2138421385  An execution thread within a `session`. Each session has one primary thread plus zero or more child threads spawned by the coordinator.2138621387  - `type: "session_thread"`2138821389  - `id: string`2139021391    Unique identifier for this thread.2139221393  - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`2139421395    The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.2139621397    - `BetaManagedAgentsSessionThreadAgent object`2139821399      Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.2140021401      - `type: "agent"`2140221403      - `id: string`2140421405      - `description: string or null`2140621407      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`2140821409        - `type: "url"`2141021411        - `name: string`2141221413        - `url: string`2141421415      - `model: BetaManagedAgentsModelConfig`2141621417        Model identifier and configuration.2141821419        - `id: BetaManagedAgentsModel`2142021421          The model that will power your agent.2142221423          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2142421425          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`2142621427            The model that will power your agent.2142821429            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2143021431            - `"claude-fable-5-1"`2143221433              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows2143421435            - `"claude-sonnet-5"`2143621437              High-performance model for coding and agents2143821439            - `"claude-fable-5"`2144021441              Next generation of intelligence for the hardest knowledge work and coding problems2144221443            - `"claude-opus-5"`2144421445              Powerful intelligence for long-running agents and coding2144621447            - `"claude-opus-4-8"`2144821449              Powerful intelligence for long-running agents and coding2145021451            - `"claude-opus-4-7"`2145221453              Powerful intelligence for long-running agents and coding2145421455            - `"claude-opus-4-6"`2145621457              Powerful intelligence for long-running agents and coding2145821459            - `"claude-sonnet-4-6"`2146021461              Best combination of speed and intelligence2146221463            - `"claude-haiku-4-5"`2146421465              Fastest model with near-frontier intelligence2146621467            - `"claude-haiku-4-5-20251001"`2146821469              Fastest model with near-frontier intelligence2147021471            - `"claude-opus-4-5"`2147221473              Powerful intelligence for long-running agents and coding2147421475            - `"claude-opus-4-5-20251101"`2147621477              Powerful intelligence for long-running agents and coding2147821479            - `"claude-sonnet-4-5"`2148021481              High-performance model for agents and coding2148221483            - `"claude-sonnet-4-5-20250929"`2148421485              High-performance model for agents and coding2148621487          - `string`2148821489        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`2149021491          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.2149221493          - `BetaManagedAgentsEffortLow object`2149421495            Low effort. Favors latency over reasoning depth.2149621497            - `type: "low"`2149821499          - `BetaManagedAgentsEffortMedium object`2150021501            Medium effort. Balances latency and reasoning depth.2150221503            - `type: "medium"`2150421505          - `BetaManagedAgentsEffortHigh object`2150621507            High effort. Favors reasoning depth.2150821509            - `type: "high"`2151021511          - `BetaManagedAgentsEffortXhigh object`2151221513            Extra-high effort. Not all models accept this level.2151421515            - `type: "xhigh"`2151621517          - `BetaManagedAgentsEffortMax object`2151821519            Maximum effort. Favors reasoning depth over latency.2152021521            - `type: "max"`2152221523        - `inference_geo: optional string`2152421525          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.2152621527        - `speed: optional "standard" or "fast"`2152821529          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.2153021531          - `"standard"`2153221533          - `"fast"`2153421535      - `name: string`2153621537      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`2153821539        - `BetaManagedAgentsAnthropicSkill object`2154021541          A resolved Anthropic-managed skill.2154221543          - `type: "anthropic"`2154421545          - `skill_id: string`2154621547          - `version: string`2154821549        - `BetaManagedAgentsCustomSkill object`2155021551          A resolved user-created custom skill.2155221553          - `type: "custom"`2155421555          - `skill_id: string`2155621557          - `version: string`2155821559      - `system: string or null`2156021561      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`2156221563        - `BetaManagedAgentsAgentToolset20260401 object`2156421565          - `type: "agent_toolset_20260401"`2156621567          - `configs: array of BetaManagedAgentsAgentToolConfig`2156821569            - `BetaManagedAgentsBashToolConfig object`2157021571              Configuration for the bash tool.2157221573              - `type: "bash"`2157421575              - `enabled: boolean`2157621577              - `name: "bash"`2157821579              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2158021581                Permission policy for tool execution.2158221583                - `BetaManagedAgentsAlwaysAllowPolicy object`2158421585                  Tool calls are automatically approved without user confirmation.2158621587                  - `type: "always_allow"`2158821589                - `BetaManagedAgentsAlwaysAskPolicy object`2159021591                  Tool calls require user confirmation before execution.2159221593                  - `type: "always_ask"`2159421595                - `BetaManagedAgentsAutoPolicy object`2159621597                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2159821599                  - `type: "auto"`2160021601            - `BetaManagedAgentsEditToolConfig object`2160221603              Configuration for the edit tool.2160421605              - `type: "edit"`2160621607              - `enabled: boolean`2160821609              - `name: "edit"`2161021611              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2161221613                Permission policy for tool execution.2161421615                - `BetaManagedAgentsAlwaysAllowPolicy object`2161621617                  Tool calls are automatically approved without user confirmation.2161821619                - `BetaManagedAgentsAlwaysAskPolicy object`2162021621                  Tool calls require user confirmation before execution.2162221623                - `BetaManagedAgentsAutoPolicy object`2162421625                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2162621627            - `BetaManagedAgentsReadToolConfig object`2162821629              Configuration for the read tool.2163021631              - `type: "read"`2163221633              - `enabled: boolean`2163421635              - `name: "read"`2163621637              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2163821639                Permission policy for tool execution.2164021641                - `BetaManagedAgentsAlwaysAllowPolicy object`2164221643                  Tool calls are automatically approved without user confirmation.2164421645                - `BetaManagedAgentsAlwaysAskPolicy object`2164621647                  Tool calls require user confirmation before execution.2164821649                - `BetaManagedAgentsAutoPolicy object`2165021651                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2165221653            - `BetaManagedAgentsWriteToolConfig object`2165421655              Configuration for the write tool.2165621657              - `type: "write"`2165821659              - `enabled: boolean`2166021661              - `name: "write"`2166221663              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2166421665                Permission policy for tool execution.2166621667                - `BetaManagedAgentsAlwaysAllowPolicy object`2166821669                  Tool calls are automatically approved without user confirmation.2167021671                - `BetaManagedAgentsAlwaysAskPolicy object`2167221673                  Tool calls require user confirmation before execution.2167421675                - `BetaManagedAgentsAutoPolicy object`2167621677                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2167821679            - `BetaManagedAgentsGlobToolConfig object`2168021681              Configuration for the glob tool.2168221683              - `type: "glob"`2168421685              - `enabled: boolean`2168621687              - `name: "glob"`2168821689              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2169021691                Permission policy for tool execution.2169221693                - `BetaManagedAgentsAlwaysAllowPolicy object`2169421695                  Tool calls are automatically approved without user confirmation.2169621697                - `BetaManagedAgentsAlwaysAskPolicy object`2169821699                  Tool calls require user confirmation before execution.2170021701                - `BetaManagedAgentsAutoPolicy object`2170221703                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2170421705            - `BetaManagedAgentsGrepToolConfig object`2170621707              Configuration for the grep tool.2170821709              - `type: "grep"`2171021711              - `enabled: boolean`2171221713              - `name: "grep"`2171421715              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2171621717                Permission policy for tool execution.2171821719                - `BetaManagedAgentsAlwaysAllowPolicy object`2172021721                  Tool calls are automatically approved without user confirmation.2172221723                - `BetaManagedAgentsAlwaysAskPolicy object`2172421725                  Tool calls require user confirmation before execution.2172621727                - `BetaManagedAgentsAutoPolicy object`2172821729                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2173021731            - `BetaManagedAgentsWebFetchToolConfig object`2173221733              Configuration for the web_fetch tool.2173421735              - `type: "web_fetch"`2173621737              - `enabled: boolean`2173821739              - `name: "web_fetch"`2174021741              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2174221743                Permission policy for tool execution.2174421745                - `BetaManagedAgentsAlwaysAllowPolicy object`2174621747                  Tool calls are automatically approved without user confirmation.2174821749                - `BetaManagedAgentsAlwaysAskPolicy object`2175021751                  Tool calls require user confirmation before execution.2175221753                - `BetaManagedAgentsAutoPolicy object`2175421755                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2175621757              - `allowed_domains: optional array of string`2175821759              - `blocked_domains: optional array of string`2176021761              - `max_content_tokens: optional number or null`2176221763                format: int322176421765            - `BetaManagedAgentsWebSearchToolConfig object`2176621767              Configuration for the web_search tool.2176821769              - `type: "web_search"`2177021771              - `enabled: boolean`2177221773              - `name: "web_search"`2177421775              - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2177621777                Permission policy for tool execution.2177821779                - `BetaManagedAgentsAlwaysAllowPolicy object`2178021781                  Tool calls are automatically approved without user confirmation.2178221783                - `BetaManagedAgentsAlwaysAskPolicy object`2178421785                  Tool calls require user confirmation before execution.2178621787                - `BetaManagedAgentsAutoPolicy object`2178821789                  The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2179021791              - `allowed_domains: optional array of string`2179221793              - `blocked_domains: optional array of string`2179421795              - `user_location: optional BetaManagedAgentsUserLocation or null`2179621797                Approximate user location for search result localization.2179821799                - `type: "approximate"`2180021801                  Location precision. Only "approximate" is supported.2180221803                - `city: optional string or null`2180421805                  City name.2180621807                  minLength: 1, maxLength: 2552180821809                - `country: optional string or null`2181021811                  Two-letter ISO 3166-1 country code, uppercase.2181221813                - `region: optional string or null`2181421815                  Region or state name.2181621817                  minLength: 1, maxLength: 2552181821819                - `timezone: optional string or null`2182021821                  IANA timezone identifier, e.g. "America/Los_Angeles".2182221823                  minLength: 1, maxLength: 2552182421825          - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`2182621827            Resolved default configuration for agent tools.2182821829            - `enabled: boolean`2183021831            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2183221833              Permission policy for tool execution.2183421835              - `BetaManagedAgentsAlwaysAllowPolicy object`2183621837                Tool calls are automatically approved without user confirmation.2183821839              - `BetaManagedAgentsAlwaysAskPolicy object`2184021841                Tool calls require user confirmation before execution.2184221843              - `BetaManagedAgentsAutoPolicy object`2184421845                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2184621847        - `BetaManagedAgentsMCPToolset object`2184821849          - `type: "mcp_toolset"`2185021851          - `configs: array of BetaManagedAgentsMCPToolConfig`2185221853            - `enabled: boolean`2185421855            - `name: string`2185621857            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2185821859              Permission policy for tool execution.2186021861              - `BetaManagedAgentsAlwaysAllowPolicy object`2186221863                Tool calls are automatically approved without user confirmation.2186421865              - `BetaManagedAgentsAlwaysAskPolicy object`2186621867                Tool calls require user confirmation before execution.2186821869              - `BetaManagedAgentsAutoPolicy object`2187021871                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2187221873          - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`2187421875            Resolved default configuration for all tools from an MCP server.2187621877            - `enabled: boolean`2187821879            - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2188021881              Permission policy for tool execution.2188221883              - `BetaManagedAgentsAlwaysAllowPolicy object`2188421885                Tool calls are automatically approved without user confirmation.2188621887              - `BetaManagedAgentsAlwaysAskPolicy object`2188821889                Tool calls require user confirmation before execution.2189021891              - `BetaManagedAgentsAutoPolicy object`2189221893                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2189421895          - `mcp_server_name: string`2189621897        - `BetaManagedAgentsCustomTool object`2189821899          A custom tool as returned in API responses.2190021901          - `type: "custom"`2190221903          - `description: string`2190421905          - `input_schema: BetaManagedAgentsCustomToolInputSchema`2190621907            JSON Schema for custom tool input parameters.2190821909            - `type: "object"`2191021911            - `properties: optional map[unknown] or null`2191221913            - `required: optional array of string or null`2191421915          - `name: string`2191621917      - `version: number`2191821919        format: int322192021921    - `BetaManagedAgentsAdvisor object`2192221923      Platform advisor roster entry: a model the session's primary thread may consult mid-turn.2192421925      - `type: "advisor"`2192621927      - `model: string`2192821929        The advisor model id.2193021931  - `archived_at: string or null`2193221933    A timestamp in RFC 3339 format2193421935    format: date-time2193621937  - `created_at: string`2193821939    A timestamp in RFC 3339 format2194021941    format: date-time2194221943  - `parent_thread_id: string or null`2194421945    Parent thread that spawned this thread. Null for the primary thread.2194621947  - `session_id: string`2194821949    The session this thread belongs to.2195021951  - `stats: BetaManagedAgentsSessionThreadStats or null`2195221953    Timing statistics for a session thread.2195421955    - `active_seconds: optional number`2195621957      Cumulative time in seconds the thread spent actively running. Excludes idle time.2195821959      format: double2196021961    - `duration_seconds: optional number`2196221963      Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.2196421965      format: double2196621967    - `startup_seconds: optional number`2196821969      Time in seconds for the thread to begin running. Zero for child threads, which start immediately.2197021971      format: double2197221973  - `status: BetaManagedAgentsSessionThreadStatus`2197421975    SessionThreadStatus enum2197621977    - `"running"`2197821979    - `"idle"`2198021981    - `"rescheduling"`2198221983    - `"terminated"`2198421985  - `updated_at: string`2198621987    A timestamp in RFC 3339 format2198821989    format: date-time2199021991  - `usage: BetaManagedAgentsSessionThreadUsage or null`2199221993    Cumulative token usage for a session thread across all turns.2199421995    - `active_seconds: optional number`2199621997      Cumulative time in seconds this thread spent in running status. Equal to `stats.active_seconds`; surfaced here so a thread's usage carries every quantity its cost is priced on.2199821999      format: double2200022001    - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`2200222003      Prompt-cache creation token usage broken down by cache lifetime.2200422005      - `ephemeral_1h_input_tokens: optional number`2200622007        Tokens used to create 1-hour ephemeral cache entries.2200822009        format: int322201022011      - `ephemeral_5m_input_tokens: optional number`2201222013        Tokens used to create 5-minute ephemeral cache entries.2201422015        format: int322201622017    - `cache_read_input_tokens: optional number`2201822019      Total tokens read from prompt cache.2202022021      format: int322202222023    - `input_tokens: optional number`2202422025      Total input tokens consumed across all turns.2202622027      format: int322202822029    - `list_cost: optional BetaMonetaryAmount or null`2203022031      A monetary amount in a specific currency.2203222033      - `amount: string`2203422035        Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.2203622037      - `currency: BetaCurrency`2203822039        Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.2204022041    - `output_tokens: optional number`2204222043      Total output tokens generated across all turns.2204422045      format: int322204622047    - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`2204822049      Cumulative count of server-executed tool invocations, broken down by tool.2205022051      - `web_fetch_requests: optional number`2205222053        Number of server-executed web fetch requests.2205422055        format: int322205622057      - `web_search_requests: optional number`2205822059        Number of server-executed web search requests.2206022061        format: int322206222063#### Example2206422065```bash22066curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID/archive \22067    -X POST \22068    -H 'anthropic-version: 2023-06-01' \22069    -H 'anthropic-beta: managed-agents-2026-04-01' \22070    -H "X-Api-Key: $ANTHROPIC_API_KEY"22071```2207222073##### Response (200)2207422075```json22076{22077  "id": "sthr_011CZkZVWa6oIjw0rgXZpnBt",22078  "agent": {22079    "id": "agent_011CZkYqphY8vELVzwCUpqiQ",22080    "description": "A focused research subagent.",22081    "mcp_servers": [22082      {22083        "name": "example-mcp",22084        "type": "url",22085        "url": "https://example-server.modelcontextprotocol.io/sse"22086      }22087    ],22088    "model": {22089      "id": "claude-opus-5",22090      "effort": {22091        "type": "low"22092      },22093      "inference_geo": "inference_geo",22094      "speed": "standard"22095    },22096    "name": "Researcher",22097    "skills": [22098      {22099        "skill_id": "xlsx",22100        "type": "anthropic",22101        "version": "1"22102      }22103    ],22104    "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",22105    "tools": [22106      {22107        "configs": [22108          {22109            "enabled": true,22110            "name": "bash",22111            "permission_policy": {22112              "type": "always_allow"22113            },22114            "type": "bash"22115          }22116        ],22117        "default_config": {22118          "enabled": true,22119          "permission_policy": {22120            "type": "always_ask"22121          }22122        },22123        "type": "agent_toolset_20260401"22124      }22125    ],22126    "type": "agent",22127    "version": 122128  },22129  "archived_at": null,22130  "created_at": "2026-03-15T10:00:00Z",22131  "parent_thread_id": null,22132  "session_id": "sesn_011CZkZAtmR3yMPDzynEDxu7",22133  "stats": {22134    "active_seconds": 0,22135    "duration_seconds": 0,22136    "startup_seconds": 022137  },22138  "status": "idle",22139  "type": "session_thread",22140  "updated_at": "2026-03-15T10:00:00Z",22141  "usage": {22142    "active_seconds": 0,22143    "cache_creation": {22144      "ephemeral_1h_input_tokens": 0,22145      "ephemeral_5m_input_tokens": 022146    },22147    "cache_read_input_tokens": 0,22148    "input_tokens": 0,22149    "list_cost": {22150      "amount": "2500",22151      "currency": "USD"22152    },22153    "output_tokens": 0,22154    "server_tool_use": {22155      "web_fetch_requests": 0,22156      "web_search_requests": 322157    }22158  }22159}22160```2216122162## Sessions › Threads › Events2216322164### List Session Thread Events2216522166**GET** `/v1/sessions/{session_id}/threads/{thread_id}/events`2216722168List Session Thread Events2216922170#### Path parameters2217122172- `session_id: string`2217322174- `thread_id: string`2217522176#### Query parameters2217722178- `limit: optional number`2217922180  Query parameter for limit2218122182  format: int322218322184- `page: optional string`2218522186  Query parameter for page2218722188#### Headers2218922190- `"anthropic-beta": optional array of AnthropicBeta`2219122192  Optional header to specify the beta version(s) you want to use.2219322194  - `string`2219522196  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`2219722198    - `"message-batches-2024-09-24"`2219922200    - `"prompt-caching-2024-07-31"`2220122202    - `"computer-use-2024-10-22"`2220322204    - `"computer-use-2025-01-24"`2220522206    - `"pdfs-2024-09-25"`2220722208    - `"token-counting-2024-11-01"`2220922210    - `"token-efficient-tools-2025-02-19"`2221122212    - `"output-128k-2025-02-19"`2221322214    - `"files-api-2025-04-14"`2221522216    - `"mcp-client-2025-04-04"`2221722218    - `"mcp-client-2025-11-20"`2221922220    - `"dev-full-thinking-2025-05-14"`2222122222    - `"interleaved-thinking-2025-05-14"`2222322224    - `"code-execution-2025-05-22"`2222522226    - `"extended-cache-ttl-2025-04-11"`2222722228    - `"context-1m-2025-08-07"`2222922230    - `"context-management-2025-06-27"`2223122232    - `"model-context-window-exceeded-2025-08-26"`2223322234    - `"skills-2025-10-02"`2223522236    - `"fast-mode-2026-02-01"`2223722238    - `"output-300k-2026-03-24"`2223922240    - `"user-profiles-2026-03-24"`2224122242    - `"user-profiles-2026-08-18"`2224322244    - `"user-profiles-2026-09-04"`2224522246    - `"advisor-tool-2026-03-01"`2224722248    - `"managed-agents-2026-04-01"`2224922250    - `"cache-diagnosis-2026-04-07"`2225122252    - `"dreaming-2026-04-21"`2225322254    - `"thinking-token-count-2026-05-13"`2225522256    - `"server-side-fallback-2026-06-01"`2225722258    - `"server-side-fallback-2026-07-01"`2225922260    - `"fallback-credit-2026-06-01"`2226122262    - `"fallback-credit-2026-07-01"`2226322264    - `"agent-memory-2026-07-22"`2226522266    - `"mid-conversation-tool-changes-2026-07-01"`2226722268    - `"compact-2026-01-12"`2226922270    - `"computer-use-2025-11-24"`2227122272    - `"mcp-tunnels-2026-06-22"`2227322274    - `"structured-outputs-2025-11-13"`2227522276    - `"task-budgets-2026-03-13"`2227722278    - `"thinking-display-updates-2026-08-18"`2227922280    - `"ce-user-management-2026-07-13"`2228122282    - `"mid-conversation-output-config-2026-07-01"`2228322284    - `"thinking-binding-controls-2026-08-01"`2228522286    - `"mid-conversation-system-clear-at-2026-08-21"`2228722288    - `"compact-2026-09-04"`2228922290- `"anthropic-workspace-id": optional string`2229122292#### Returns2229322294- `data: optional array of BetaManagedAgentsSessionEvent`2229522296  Events for the thread, ordered by `processed_at`.2229722298  - `BetaManagedAgentsUserMessageEvent object`2229922300    A user message event in the session conversation.2230122302    - `type: "user.message"`2230322304    - `id: string`2230522306      Unique identifier for this event.2230722308    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`2230922310      Array of content blocks comprising the user message.2231122312      - `BetaManagedAgentsTextBlock object`2231322314        Regular text content.2231522316        - `type: "text"`2231722318        - `text: string`2231922320          The text content.2232122322          minLength: 12232322324      - `BetaManagedAgentsImageBlock object`2232522326        Image content specified directly as base64 data or as a reference via a URL.2232722328        - `type: "image"`2232922330        - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`2233122332          Union type for image source variants.2233322334          - `BetaManagedAgentsBase64ImageSource object`2233522336            Base64-encoded image data.2233722338            - `type: "base64"`2233922340            - `data: string`2234122342              Base64-encoded image data.2234322344              minLength: 12234522346            - `media_type: string`2234722348              MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").2234922350              minLength: 12235122352          - `BetaManagedAgentsURLImageSource object`2235322354            Image referenced by URL.2235522356            - `type: "url"`2235722358            - `url: string`2235922360              URL of the image to fetch.2236122362              minLength: 12236322364          - `BetaManagedAgentsFileImageSource object`2236522366            Image referenced by file ID.2236722368            - `type: "file"`2236922370            - `file_id: string`2237122372              ID of a previously uploaded file.2237322374              minLength: 12237522376      - `BetaManagedAgentsDocumentBlock object`2237722378        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2237922380        - `type: "document"`2238122382        - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`2238322384          Union type for document source variants.2238522386          - `BetaManagedAgentsBase64DocumentSource object`2238722388            Base64-encoded document data.2238922390            - `type: "base64"`2239122392            - `data: string`2239322394              Base64-encoded document data.2239522396              minLength: 12239722398            - `media_type: string`2239922400              MIME type of the document (e.g., "application/pdf").2240122402              minLength: 12240322404          - `BetaManagedAgentsPlainTextDocumentSource object`2240522406            Plain text document content.2240722408            - `type: "text"`2240922410            - `data: string`2241122412              The plain text content.2241322414              minLength: 12241522416            - `media_type: "text/plain"`2241722418              MIME type of the text content. Must be "text/plain".2241922420          - `BetaManagedAgentsURLDocumentSource object`2242122422            Document referenced by URL.2242322424            - `type: "url"`2242522426            - `url: string`2242722428              URL of the document to fetch.2242922430              minLength: 12243122432          - `BetaManagedAgentsFileDocumentSource object`2243322434            Document referenced by file ID.2243522436            - `type: "file"`2243722438            - `file_id: string`2243922440              ID of a previously uploaded file.2244122442              minLength: 12244322444        - `context: optional string or null`2244522446          Additional context about the document for the model.2244722448        - `title: optional string or null`2244922450          The title of the document.2245122452      - `BetaManagedAgentsRedactedBlock object`2245322454        Placeholder for content withheld by Anthropic model policy.2245522456        - `type: "redacted"`2245722458    - `processed_at: optional string or null`2245922460      A timestamp in RFC 3339 format2246122462      format: date-time2246322464  - `BetaManagedAgentsUserInterruptEvent object`2246522466    An interrupt event that pauses agent execution and returns control to the user.2246722468    - `type: "user.interrupt"`2246922470    - `id: string`2247122472      Unique identifier for this event.2247322474    - `processed_at: optional string or null`2247522476      A timestamp in RFC 3339 format2247722478      format: date-time2247922480    - `session_thread_id: optional string or null`2248122482      If absent, interrupts every non-archived thread in a multiagent session (or the primary alone in a single-agent session). If present, interrupts only the named thread.2248322484  - `BetaManagedAgentsUserToolConfirmationEvent object`2248522486    A tool confirmation event that approves or denies a pending tool execution.2248722488    - `type: "user.tool_confirmation"`2248922490    - `id: string`2249122492      Unique identifier for this event.2249322494    - `result: "allow" or "deny"`2249522496      UserToolConfirmationResult enum2249722498      - `"allow"`2249922500      - `"deny"`2250122502    - `tool_use_id: string`2250322504      The id of the `agent.tool_use` or `agent.mcp_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.2250522506    - `deny_message: optional string or null`2250722508      Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.2250922510      maxLength: 100002251122512    - `processed_at: optional string or null`2251322514      A timestamp in RFC 3339 format2251522516      format: date-time2251722518    - `session_thread_id: optional string or null`2251922520      Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.2252122522  - `BetaManagedAgentsUserCustomToolResultEvent object`2252322524    Event sent by the client providing the result of a custom tool execution.2252522526    - `type: "user.custom_tool_result"`2252722528    - `id: string`2252922530      Unique identifier for this event.2253122532    - `custom_tool_use_id: string`2253322534      The id of the `agent.custom_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.2253522536    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2253722538      The result content returned by the tool.2253922540      - `BetaManagedAgentsTextBlock object`2254122542        Regular text content.2254322544      - `BetaManagedAgentsImageBlock object`2254522546        Image content specified directly as base64 data or as a reference via a URL.2254722548      - `BetaManagedAgentsDocumentBlock object`2254922550        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2255122552      - `BetaManagedAgentsSearchResultBlock object`2255322554        A block containing a web search result.2255522556        - `type: "search_result"`2255722558        - `citations: BetaManagedAgentsSearchResultCitations`2255922560          Citation settings for a search result.2256122562          - `enabled: boolean`2256322564            Whether citations are enabled for this search result.2256522566        - `content: array of BetaManagedAgentsSearchResultContent`2256722568          Array of text content blocks from the search result.2256922570          - `type: "text"`2257122572          - `text: string`2257322574            The text content.2257522576            minLength: 12257722578        - `source: string`2257922580          The URL source of the search result.2258122582          minLength: 12258322584        - `title: string`2258522586          The title of the search result.2258722588          minLength: 12258922590    - `is_error: optional boolean or null`2259122592      Whether the tool execution resulted in an error.2259322594    - `processed_at: optional string or null`2259522596      A timestamp in RFC 3339 format2259722598      format: date-time2259922600    - `session_thread_id: optional string or null`2260122602      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.2260322604  - `BetaManagedAgentsAgentCustomToolUseEvent object`2260522606    Event emitted when the agent calls a custom tool. The session goes idle until the client sends a `user.custom_tool_result` event with the result.2260722608    - `type: "agent.custom_tool_use"`2260922610    - `id: string`2261122612      Unique identifier for this event.2261322614    - `input: map[unknown]`2261522616      Input parameters for the tool call.2261722618    - `name: string`2261922620      Name of the custom tool being called.2262122622    - `processed_at: string`2262322624      A timestamp in RFC 3339 format2262522626      format: date-time2262722628    - `session_thread_id: optional string or null`2262922630      When set, this event was cross-posted from a subagent's thread to surface its custom tool use on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.custom_tool_result` by `custom_tool_use_id`, so clients do not send it back.2263122632  - `BetaManagedAgentsAgentMessageEvent object`2263322634    An agent response event in the session conversation.2263522636    - `type: "agent.message"`2263722638    - `id: string`2263922640      Unique identifier for this event.2264122642    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`2264322644      Array of text blocks comprising the agent response.2264522646      - `BetaManagedAgentsTextBlock object`2264722648        Regular text content.2264922650      - `BetaManagedAgentsRedactedBlock object`2265122652        Placeholder for content withheld by Anthropic model policy.2265322654    - `processed_at: string`2265522656      A timestamp in RFC 3339 format2265722658      format: date-time2265922660  - `BetaManagedAgentsAgentThinkingEvent object`2266122662    Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.2266322664    - `type: "agent.thinking"`2266522666    - `id: string`2266722668      Unique identifier for this event.2266922670    - `processed_at: string`2267122672      A timestamp in RFC 3339 format2267322674      format: date-time2267522676  - `BetaManagedAgentsAgentMCPToolUseEvent object`2267722678    Event emitted when the agent invokes a tool provided by an MCP server.2267922680    - `type: "agent.mcp_tool_use"`2268122682    - `id: string`2268322684      Unique identifier for this event.2268522686    - `input: map[unknown]`2268722688      Input parameters for the tool call.2268922690    - `mcp_server_name: string`2269122692      Name of the MCP server providing the tool.2269322694    - `name: string`2269522696      Name of the MCP tool being used.2269722698    - `processed_at: string`2269922700      A timestamp in RFC 3339 format2270122702      format: date-time2270322704    - `evaluated_permission: optional "allow" or "ask" or "deny"`2270522706      AgentEvaluatedPermission enum2270722708      - `"allow"`2270922710      - `"ask"`2271122712      - `"deny"`2271322714    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`2271522716      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.2271722718      - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`2271922720        The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".2272122722        - `type: "always_allow"`2272322724      - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`2272522726        The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".2272722728        - `type: "always_ask"`2272922730      - `BetaManagedAgentsAgentToolEvaluationAuto object`2273122732        The resolved permission_policy was auto: the server judged this invocation individually.2273322734        - `type: "auto"`2273522736        - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`2273722738          The server's per-invocation judgement under the auto permission policy. Its type always equals the event's top-level evaluated_permission. Open union: clients must tolerate unknown variants.2273922740          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`2274122742            The server judged the invocation safe to execute without client approval.2274322744            - `type: "allow"`2274522746          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`2274722748            The server reached no judgement; the invocation is held for client approval.2274922750            - `type: "ask"`2275122752            - `reason_code: string`2275322754              The judgement's grounds in registry-bound terms, for client branching and audit rather than end-user display. Open registry; currently "indeterminate" (no judgement was reached). Clients must tolerate values outside this set.2275522756              maxLength: 642275722758          - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`2275922760            The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.2276122762            - `type: "deny"`2276322764            - `reason_code: string`2276522766              The judgement's grounds in registry-bound terms. Open registry; currently "high_risk" (judged high-risk; the call does not run). Clients must tolerate values outside this set.2276722768              maxLength: 642276922770    - `session_thread_id: optional string or null`2277122772      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` by `tool_use_id`, so clients do not send it back.2277322774  - `BetaManagedAgentsAgentMCPToolResultEvent object`2277522776    Event representing the result of an MCP tool execution.2277722778    - `type: "agent.mcp_tool_result"`2277922780    - `id: string`2278122782      Unique identifier for this event.2278322784    - `mcp_tool_use_id: string`2278522786      The id of the `agent.mcp_tool_use` event this result corresponds to.2278722788    - `processed_at: string`2278922790      A timestamp in RFC 3339 format2279122792      format: date-time2279322794    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2279522796      The result content returned by the tool.2279722798      - `BetaManagedAgentsTextBlock object`2279922800        Regular text content.2280122802      - `BetaManagedAgentsImageBlock object`2280322804        Image content specified directly as base64 data or as a reference via a URL.2280522806      - `BetaManagedAgentsDocumentBlock object`2280722808        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2280922810      - `BetaManagedAgentsSearchResultBlock object`2281122812        A block containing a web search result.2281322814    - `is_error: optional boolean or null`2281522816      Whether the tool execution resulted in an error.2281722818  - `BetaManagedAgentsAgentToolUseEvent object`2281922820    Event emitted when the agent invokes a built-in agent tool.2282122822    - `type: "agent.tool_use"`2282322824    - `id: string`2282522826      Unique identifier for this event.2282722828    - `input: map[unknown]`2282922830      Input parameters for the tool call.2283122832    - `name: string`2283322834      Name of the agent tool being used.2283522836    - `processed_at: string`2283722838      A timestamp in RFC 3339 format2283922840      format: date-time2284122842    - `evaluated_permission: optional "allow" or "ask" or "deny"`2284322844      AgentEvaluatedPermission enum2284522846      - `"allow"`2284722848      - `"ask"`2284922850      - `"deny"`2285122852    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`2285322854      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.2285522856    - `session_thread_id: optional string or null`2285722858      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` or `user.tool_result` by `tool_use_id`, so clients do not send it back.2285922860  - `BetaManagedAgentsAgentToolResultEvent object`2286122862    Event representing the result of an agent tool execution.2286322864    - `type: "agent.tool_result"`2286522866    - `id: string`2286722868      Unique identifier for this event.2286922870    - `processed_at: string`2287122872      A timestamp in RFC 3339 format2287322874      format: date-time2287522876    - `tool_use_id: string`2287722878      The id of the `agent.tool_use` event this result corresponds to.2287922880    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2288122882      The result content returned by the tool.2288322884      - `BetaManagedAgentsTextBlock object`2288522886        Regular text content.2288722888      - `BetaManagedAgentsImageBlock object`2288922890        Image content specified directly as base64 data or as a reference via a URL.2289122892      - `BetaManagedAgentsDocumentBlock object`2289322894        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2289522896      - `BetaManagedAgentsSearchResultBlock object`2289722898        A block containing a web search result.2289922900    - `is_error: optional boolean or null`2290122902      Whether the tool execution resulted in an error.2290322904  - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`2290522906    Delivery event written to the target thread's input stream when an agent-to-agent message arrives.2290722908    - `type: "agent.thread_message_received"`2290922910    - `id: string`2291122912      Unique identifier for this event.2291322914    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`2291522916      Message content blocks.2291722918      - `BetaManagedAgentsTextBlock object`2291922920        Regular text content.2292122922      - `BetaManagedAgentsImageBlock object`2292322924        Image content specified directly as base64 data or as a reference via a URL.2292522926      - `BetaManagedAgentsDocumentBlock object`2292722928        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2292922930      - `BetaManagedAgentsRedactedBlock object`2293122932        Placeholder for content withheld by Anthropic model policy.2293322934    - `from_session_thread_id: string`2293522936      Public `sthr_` ID of the thread that sent the message.2293722938    - `processed_at: string`2293922940      A timestamp in RFC 3339 format2294122942      format: date-time2294322944    - `from_agent_name: optional string or null`2294522946      Name of the callable agent this message came from. Absent when received from the primary agent.2294722948  - `BetaManagedAgentsAgentThreadMessageSentEvent object`2294922950    Observability event emitted to the sender's output stream when an agent-to-agent message is sent.2295122952    - `type: "agent.thread_message_sent"`2295322954    - `id: string`2295522956      Unique identifier for this event.2295722958    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`2295922960      Message content blocks.2296122962      - `BetaManagedAgentsTextBlock object`2296322964        Regular text content.2296522966      - `BetaManagedAgentsImageBlock object`2296722968        Image content specified directly as base64 data or as a reference via a URL.2296922970      - `BetaManagedAgentsDocumentBlock object`2297122972        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2297322974      - `BetaManagedAgentsRedactedBlock object`2297522976        Placeholder for content withheld by Anthropic model policy.2297722978    - `processed_at: string`2297922980      A timestamp in RFC 3339 format2298122982      format: date-time2298322984    - `to_session_thread_id: string`2298522986      Public `sthr_` ID of the thread the message was sent to.2298722988    - `to_agent_name: optional string or null`2298922990      Name of the callable agent this message was sent to. Absent when sent to the primary agent.2299122992  - `BetaManagedAgentsAgentThreadContextCompactedEvent object`2299322994    Indicates that context compaction (summarization) occurred during the session.2299522996    - `type: "agent.thread_context_compacted"`2299722998    - `id: string`2299923000      Unique identifier for this event.2300123002    - `processed_at: string`2300323004      A timestamp in RFC 3339 format2300523006      format: date-time2300723008  - `BetaManagedAgentsSessionErrorEvent object`2300923010    An error event indicating a problem occurred during session execution.2301123012    - `type: "session.error"`2301323014    - `id: string`2301523016      Unique identifier for this event.2301723018    - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`2301923020      - `BetaManagedAgentsUnknownError object`2302123022        An unknown or unexpected error occurred during session execution. A fallback variant; clients that don't recognize a new error code can match on `retry_status` and `message` alone.2302323024        - `type: "unknown_error"`2302523026        - `message: string`2302723028          Human-readable error description.2302923030        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2303123032          What the client should do next in response to this error.2303323034          - `BetaManagedAgentsRetryStatusRetrying object`2303523036            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2303723038            - `type: "retrying"`2303923040          - `BetaManagedAgentsRetryStatusExhausted object`2304123042            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2304323044            - `type: "exhausted"`2304523046          - `BetaManagedAgentsRetryStatusTerminal object`2304723048            The session encountered a terminal error and will transition to `terminated` state.2304923050            - `type: "terminal"`2305123052      - `BetaManagedAgentsModelOverloadedError object`2305323054        The model is currently overloaded. Emitted after automatic retries are exhausted.2305523056        - `type: "model_overloaded_error"`2305723058        - `message: string`2305923060          Human-readable error description.2306123062        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2306323064          What the client should do next in response to this error.2306523066          - `BetaManagedAgentsRetryStatusRetrying object`2306723068            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2306923070          - `BetaManagedAgentsRetryStatusExhausted object`2307123072            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2307323074          - `BetaManagedAgentsRetryStatusTerminal object`2307523076            The session encountered a terminal error and will transition to `terminated` state.2307723078      - `BetaManagedAgentsModelRateLimitedError object`2307923080        The model request was rate-limited.2308123082        - `type: "model_rate_limited_error"`2308323084        - `message: string`2308523086          Human-readable error description.2308723088        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2308923090          What the client should do next in response to this error.2309123092          - `BetaManagedAgentsRetryStatusRetrying object`2309323094            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2309523096          - `BetaManagedAgentsRetryStatusExhausted object`2309723098            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2309923100          - `BetaManagedAgentsRetryStatusTerminal object`2310123102            The session encountered a terminal error and will transition to `terminated` state.2310323104      - `BetaManagedAgentsModelRequestFailedError object`2310523106        A model request failed for a reason other than overload or rate-limiting.2310723108        - `type: "model_request_failed_error"`2310923110        - `message: string`2311123112          Human-readable error description.2311323114        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2311523116          What the client should do next in response to this error.2311723118          - `BetaManagedAgentsRetryStatusRetrying object`2311923120            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2312123122          - `BetaManagedAgentsRetryStatusExhausted object`2312323124            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2312523126          - `BetaManagedAgentsRetryStatusTerminal object`2312723128            The session encountered a terminal error and will transition to `terminated` state.2312923130      - `BetaManagedAgentsMCPConnectionFailedError object`2313123132        Failed to connect to an MCP server.2313323134        - `type: "mcp_connection_failed_error"`2313523136        - `mcp_server_name: string`2313723138          Name of the MCP server that failed to connect.2313923140        - `message: string`2314123142          Human-readable error description.2314323144        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2314523146          What the client should do next in response to this error.2314723148          - `BetaManagedAgentsRetryStatusRetrying object`2314923150            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2315123152          - `BetaManagedAgentsRetryStatusExhausted object`2315323154            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2315523156          - `BetaManagedAgentsRetryStatusTerminal object`2315723158            The session encountered a terminal error and will transition to `terminated` state.2315923160      - `BetaManagedAgentsMCPAuthenticationFailedError object`2316123162        Authentication to an MCP server failed.2316323164        - `type: "mcp_authentication_failed_error"`2316523166        - `mcp_server_name: string`2316723168          Name of the MCP server that failed authentication.2316923170        - `message: string`2317123172          Human-readable error description.2317323174        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2317523176          What the client should do next in response to this error.2317723178          - `BetaManagedAgentsRetryStatusRetrying object`2317923180            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2318123182          - `BetaManagedAgentsRetryStatusExhausted object`2318323184            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2318523186          - `BetaManagedAgentsRetryStatusTerminal object`2318723188            The session encountered a terminal error and will transition to `terminated` state.2318923190      - `BetaManagedAgentsBillingError object`2319123192        The caller's organization or workspace cannot make model requests — out of credits or spend limit reached. Retrying with the same credentials will not succeed; the caller must resolve the billing state.2319323194        - `type: "billing_error"`2319523196        - `message: string`2319723198          Human-readable error description.2319923200        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2320123202          What the client should do next in response to this error.2320323204          - `BetaManagedAgentsRetryStatusRetrying object`2320523206            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2320723208          - `BetaManagedAgentsRetryStatusExhausted object`2320923210            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2321123212          - `BetaManagedAgentsRetryStatusTerminal object`2321323214            The session encountered a terminal error and will transition to `terminated` state.2321523216      - `BetaManagedAgentsCredentialHostUnreachableError object`2321723218        An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.2321923220        - `type: "credential_host_unreachable_error"`2322123222        - `credential_id: string`2322323224          ID of the affected credential.2322523226        - `message: string`2322723228          Human-readable error description.2322923230        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2323123232          What the client should do next in response to this error.2323323234          - `BetaManagedAgentsRetryStatusRetrying object`2323523236            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2323723238          - `BetaManagedAgentsRetryStatusExhausted object`2323923240            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2324123242          - `BetaManagedAgentsRetryStatusTerminal object`2324323244            The session encountered a terminal error and will transition to `terminated` state.2324523246        - `vault_id: string`2324723248          ID of the vault containing the affected credential.2324923250    - `processed_at: string`2325123252      A timestamp in RFC 3339 format2325323254      format: date-time2325523256  - `BetaManagedAgentsSessionStatusRescheduledEvent object`2325723258    Indicates the session is recovering from an error state and is rescheduled for execution.2325923260    - `type: "session.status_rescheduled"`2326123262    - `id: string`2326323264      Unique identifier for this event.2326523266    - `processed_at: string`2326723268      A timestamp in RFC 3339 format2326923270      format: date-time2327123272  - `BetaManagedAgentsSessionStatusRunningEvent object`2327323274    Indicates the session is actively running and the agent is working.2327523276    - `type: "session.status_running"`2327723278    - `id: string`2327923280      Unique identifier for this event.2328123282    - `processed_at: string`2328323284      A timestamp in RFC 3339 format2328523286      format: date-time2328723288  - `BetaManagedAgentsSessionStatusIdleEvent object`2328923290    Indicates the agent has paused and is awaiting user input.2329123292    - `type: "session.status_idle"`2329323294    - `id: string`2329523296      Unique identifier for this event.2329723298    - `processed_at: string`2329923300      A timestamp in RFC 3339 format2330123302      format: date-time2330323304    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`2330523306      - `BetaManagedAgentsSessionEndTurn object`2330723308        The agent completed its turn naturally and is ready for the next user message.2330923310        - `type: "end_turn"`2331123312      - `BetaManagedAgentsSessionRequiresAction object`2331323314        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.2331523316        - `type: "requires_action"`2331723318        - `event_ids: array of string`2331923320          The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.2332123322      - `BetaManagedAgentsSessionRetriesExhausted object`2332323324        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.2332523326        - `type: "retries_exhausted"`2332723328      - `BetaManagedAgentsSessionBudgetReached object`2332923330        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.2333123332        - `type: "budget_reached"`2333323334  - `BetaManagedAgentsSessionStatusTerminatedEvent object`2333523336    Indicates the session has terminated, either due to an error or completion.2333723338    - `type: "session.status_terminated"`2333923340    - `id: string`2334123342      Unique identifier for this event.2334323344    - `processed_at: string`2334523346      A timestamp in RFC 3339 format2334723348      format: date-time2334923350  - `BetaManagedAgentsSessionThreadCreatedEvent object`2335123352    Emitted when a subagent is spawned as a new thread. Written to the parent thread's output stream so clients observing the session see child creation.2335323354    - `type: "session.thread_created"`2335523356    - `id: string`2335723358      Unique identifier for this event.2335923360    - `agent_name: string`2336123362      Name of the callable agent the thread runs.2336323364    - `processed_at: string`2336523366      A timestamp in RFC 3339 format2336723368      format: date-time2336923370    - `session_thread_id: string`2337123372      Public `sthr_` ID of the newly created thread.2337323374  - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`2337523376    Emitted when an outcome evaluation cycle begins.2337723378    - `type: "span.outcome_evaluation_start"`2337923380    - `id: string`2338123382      Unique identifier for this event.2338323384    - `iteration: number`2338523386      0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.2338723388      format: int322338923390    - `outcome_id: string`2339123392      The `outc_` ID of the outcome being evaluated.2339323394    - `processed_at: string`2339523396      A timestamp in RFC 3339 format2339723398      format: date-time2339923400  - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`2340123402    Emitted when an outcome evaluation cycle completes. Carries the verdict and aggregate token usage. A verdict of `needs_revision` means another evaluation cycle follows; `satisfied`, `max_iterations_reached`, `failed`, or `interrupted` are terminal — no further evaluation cycles follow.2340323404    - `type: "span.outcome_evaluation_end"`2340523406    - `id: string`2340723408      Unique identifier for this event.2340923410    - `explanation: string`2341123412      Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.2341323414    - `iteration: number`2341523416      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.2341723418      format: int322341923420    - `outcome_evaluation_start_id: string`2342123422      The id of the corresponding `span.outcome_evaluation_start` event.2342323424    - `outcome_id: string`2342523426      The `outc_` ID of the outcome being evaluated.2342723428    - `processed_at: string`2342923430      A timestamp in RFC 3339 format2343123432      format: date-time2343323434    - `result: string`2343523436      Evaluation verdict. 'satisfied': criteria met, session goes idle. 'needs_revision': criteria not met, another revision cycle follows. 'max_iterations_reached': evaluation budget exhausted with criteria still unmet — one final acknowledgment turn follows before the session goes idle, but no further evaluation runs. 'failed': grader determined the rubric does not apply to the deliverables. 'interrupted': user sent an interrupt while evaluation was in progress.2343723438    - `usage: BetaManagedAgentsSpanModelUsage`2343923440      Token usage for a single model request.2344123442      - `cache_creation_input_tokens: number`2344323444        Tokens used to create prompt cache in this request.2344523446        format: int322344723448      - `cache_read_input_tokens: number`2344923450        Tokens read from prompt cache in this request.2345123452        format: int322345323454      - `input_tokens: number`2345523456        Input tokens consumed by this request.2345723458        format: int322345923460      - `output_tokens: number`2346123462        Output tokens generated by this request.2346323464        format: int322346523466      - `speed: optional "standard" or "fast" or null`2346723468        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.2346923470        - `"standard"`2347123472        - `"fast"`2347323474  - `BetaManagedAgentsSpanModelRequestStartEvent object`2347523476    Emitted when a model request is initiated by the agent.2347723478    - `type: "span.model_request_start"`2347923480    - `id: string`2348123482      Unique identifier for this event.2348323484    - `processed_at: string`2348523486      A timestamp in RFC 3339 format2348723488      format: date-time2348923490  - `BetaManagedAgentsSpanModelRequestEndEvent object`2349123492    Emitted when a model request completes.2349323494    - `type: "span.model_request_end"`2349523496    - `id: string`2349723498      Unique identifier for this event.2349923500    - `is_error: boolean or null`2350123502      Whether the model request resulted in an error.2350323504    - `model_request_start_id: string`2350523506      The id of the corresponding `span.model_request_start` event.2350723508    - `model_usage: BetaManagedAgentsSpanModelUsage`2350923510      Token usage for a single model request.2351123512    - `processed_at: string`2351323514      A timestamp in RFC 3339 format2351523516      format: date-time2351723518  - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`2351923520    Periodic heartbeat emitted while an outcome evaluation cycle is in progress. Distinguishes 'evaluation is actively running' from 'evaluation is stuck' between the corresponding `span.outcome_evaluation_start` and `span.outcome_evaluation_end` events.2352123522    - `type: "span.outcome_evaluation_ongoing"`2352323524    - `id: string`2352523526      Unique identifier for this event.2352723528    - `iteration: number`2352923530      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.2353123532      format: int322353323534    - `outcome_id: string`2353523536      The `outc_` ID of the outcome being evaluated.2353723538    - `processed_at: string`2353923540      A timestamp in RFC 3339 format2354123542      format: date-time2354323544  - `BetaManagedAgentsUserDefineOutcomeEvent object`2354523546    Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.2354723548    - `type: "user.define_outcome"`2354923550    - `id: string`2355123552      Unique identifier for this event.2355323554    - `description: string`2355523556      What the agent should produce. Copied from the input event.2355723558    - `max_iterations: number or null`2355923560      Evaluate-then-revise cycles before giving up. Default 3, max 20.2356123562      format: int322356323564    - `outcome_id: string`2356523566      Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.2356723568    - `processed_at: string`2356923570      A timestamp in RFC 3339 format2357123572      format: date-time2357323574    - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`2357523576      Rubric for grading the quality of an outcome.2357723578      - `BetaManagedAgentsFileRubric object`2357923580        Rubric referenced by a file uploaded via the Files API.2358123582        - `type: "file"`2358323584        - `file_id: string`2358523586          ID of the rubric file.2358723588      - `BetaManagedAgentsTextRubric object`2358923590        Rubric content provided inline as text.2359123592        - `type: "text"`2359323594        - `content: string`2359523596          Rubric content. Plain text or markdown — the grader treats it as freeform text.2359723598  - `BetaManagedAgentsSessionDeletedEvent object`2359923600    Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.2360123602    - `type: "session.deleted"`2360323604    - `id: string`2360523606      Unique identifier for this event.2360723608    - `processed_at: string`2360923610      A timestamp in RFC 3339 format2361123612      format: date-time2361323614  - `BetaManagedAgentsSessionThreadStatusRunningEvent object`2361523616    A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2361723618    - `type: "session.thread_status_running"`2361923620    - `id: string`2362123622      Unique identifier for this event.2362323624    - `agent_name: string`2362523626      Name of the agent the thread runs.2362723628    - `processed_at: string`2362923630      A timestamp in RFC 3339 format2363123632      format: date-time2363323634    - `session_thread_id: string`2363523636      Public sthr_ ID of the thread that started running.2363723638  - `BetaManagedAgentsSessionThreadStatusIdleEvent object`2363923640    A session thread has yielded and is awaiting input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2364123642    - `type: "session.thread_status_idle"`2364323644    - `id: string`2364523646      Unique identifier for this event.2364723648    - `agent_name: string`2364923650      Name of the agent the thread runs.2365123652    - `processed_at: string`2365323654      A timestamp in RFC 3339 format2365523656      format: date-time2365723658    - `session_thread_id: string`2365923660      Public sthr_ ID of the thread that went idle.2366123662    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`2366323664      - `BetaManagedAgentsSessionEndTurn object`2366523666        The agent completed its turn naturally and is ready for the next user message.2366723668      - `BetaManagedAgentsSessionRequiresAction object`2366923670        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.2367123672      - `BetaManagedAgentsSessionRetriesExhausted object`2367323674        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.2367523676      - `BetaManagedAgentsSessionBudgetReached object`2367723678        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.2367923680  - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`2368123682    A session thread has terminated and will accept no further input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2368323684    - `type: "session.thread_status_terminated"`2368523686    - `id: string`2368723688      Unique identifier for this event.2368923690    - `agent_name: string`2369123692      Name of the agent the thread runs.2369323694    - `processed_at: string`2369523696      A timestamp in RFC 3339 format2369723698      format: date-time2369923700    - `session_thread_id: string`2370123702      Public sthr_ ID of the thread that terminated.2370323704  - `BetaManagedAgentsUserToolResultEvent object`2370523706    Event sent by the client providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.2370723708    - `type: "user.tool_result"`2370923710    - `id: string`2371123712      Unique identifier for this event.2371323714    - `tool_use_id: string`2371523716      The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.2371723718    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2371923720      The result content returned by the tool.2372123722      - `BetaManagedAgentsTextBlock object`2372323724        Regular text content.2372523726      - `BetaManagedAgentsImageBlock object`2372723728        Image content specified directly as base64 data or as a reference via a URL.2372923730      - `BetaManagedAgentsDocumentBlock object`2373123732        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2373323734      - `BetaManagedAgentsSearchResultBlock object`2373523736        A block containing a web search result.2373723738    - `is_error: optional boolean or null`2373923740      Whether the tool execution resulted in an error.2374123742    - `processed_at: optional string or null`2374323744      A timestamp in RFC 3339 format2374523746      format: date-time2374723748    - `session_thread_id: optional string or null`2374923750      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.2375123752  - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`2375323754    A session thread hit a transient error and is retrying automatically. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2375523756    - `type: "session.thread_status_rescheduled"`2375723758    - `id: string`2375923760      Unique identifier for this event.2376123762    - `agent_name: string`2376323764      Name of the agent the thread runs.2376523766    - `processed_at: string`2376723768      A timestamp in RFC 3339 format2376923770      format: date-time2377123772    - `session_thread_id: string`2377323774      Public sthr_ ID of the thread that is retrying.2377523776  - `BetaManagedAgentsSessionUpdatedEvent object`2377723778    Emitted when an UpdateSession request changed at least one field. Carries only the fields that changed; absent fields were not part of the update. The new configuration applies from the next turn.2377923780    - `type: "session.updated"`2378123782    - `id: string`2378323784      Unique identifier for this event.2378523786    - `processed_at: string`2378723788      A timestamp in RFC 3339 format2378923790      format: date-time2379123792    - `agent: optional BetaManagedAgentsSessionAgent or null`2379323794      Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.2379523796      - `type: "agent"`2379723798      - `id: string`2379923800      - `description: string or null`2380123802      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`2380323804        - `type: "url"`2380523806        - `name: string`2380723808        - `url: string`2380923810      - `model: BetaManagedAgentsModelConfig`2381123812        Model identifier and configuration.2381323814        - `id: BetaManagedAgentsModel`2381523816          The model that will power your agent.2381723818          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2381923820          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`2382123822            The model that will power your agent.2382323824            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2382523826            - `"claude-fable-5-1"`2382723828              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows2382923830            - `"claude-sonnet-5"`2383123832              High-performance model for coding and agents2383323834            - `"claude-fable-5"`2383523836              Next generation of intelligence for the hardest knowledge work and coding problems2383723838            - `"claude-opus-5"`2383923840              Powerful intelligence for long-running agents and coding2384123842            - `"claude-opus-4-8"`2384323844              Powerful intelligence for long-running agents and coding2384523846            - `"claude-opus-4-7"`2384723848              Powerful intelligence for long-running agents and coding2384923850            - `"claude-opus-4-6"`2385123852              Powerful intelligence for long-running agents and coding2385323854            - `"claude-sonnet-4-6"`2385523856              Best combination of speed and intelligence2385723858            - `"claude-haiku-4-5"`2385923860              Fastest model with near-frontier intelligence2386123862            - `"claude-haiku-4-5-20251001"`2386323864              Fastest model with near-frontier intelligence2386523866            - `"claude-opus-4-5"`2386723868              Powerful intelligence for long-running agents and coding2386923870            - `"claude-opus-4-5-20251101"`2387123872              Powerful intelligence for long-running agents and coding2387323874            - `"claude-sonnet-4-5"`2387523876              High-performance model for agents and coding2387723878            - `"claude-sonnet-4-5-20250929"`2387923880              High-performance model for agents and coding2388123882          - `string`2388323884        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`2388523886          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.2388723888          - `BetaManagedAgentsEffortLow object`2388923890            Low effort. Favors latency over reasoning depth.2389123892            - `type: "low"`2389323894          - `BetaManagedAgentsEffortMedium object`2389523896            Medium effort. Balances latency and reasoning depth.2389723898            - `type: "medium"`2389923900          - `BetaManagedAgentsEffortHigh object`2390123902            High effort. Favors reasoning depth.2390323904            - `type: "high"`2390523906          - `BetaManagedAgentsEffortXhigh object`2390723908            Extra-high effort. Not all models accept this level.2390923910            - `type: "xhigh"`2391123912          - `BetaManagedAgentsEffortMax object`2391323914            Maximum effort. Favors reasoning depth over latency.2391523916            - `type: "max"`2391723918        - `inference_geo: optional string`2391923920          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.2392123922        - `speed: optional "standard" or "fast"`2392323924          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.2392523926          - `"standard"`2392723928          - `"fast"`2392923930      - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`2393123932        Resolved coordinator topology with full agent definitions for each roster member.2393323934        - `type: "coordinator"`2393523936        - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`2393723938          Full `agent` definitions the coordinator may spawn as session threads.2393923940          - `BetaManagedAgentsSessionThreadAgent object`2394123942            Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.2394323944            - `type: "agent"`2394523946            - `id: string`2394723948            - `description: string or null`2394923950            - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`2395123952              - `type: "url"`2395323954              - `name: string`2395523956              - `url: string`2395723958            - `model: BetaManagedAgentsModelConfig`2395923960              Model identifier and configuration.2396123962            - `name: string`2396323964            - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`2396523966              - `BetaManagedAgentsAnthropicSkill object`2396723968                A resolved Anthropic-managed skill.2396923970                - `type: "anthropic"`2397123972                - `skill_id: string`2397323974                - `version: string`2397523976              - `BetaManagedAgentsCustomSkill object`2397723978                A resolved user-created custom skill.2397923980                - `type: "custom"`2398123982                - `skill_id: string`2398323984                - `version: string`2398523986            - `system: string or null`2398723988            - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`2398923990              - `BetaManagedAgentsAgentToolset20260401 object`2399123992                - `type: "agent_toolset_20260401"`2399323994                - `configs: array of BetaManagedAgentsAgentToolConfig`2399523996                  - `BetaManagedAgentsBashToolConfig object`2399723998                    Configuration for the bash tool.2399924000                    - `type: "bash"`2400124002                    - `enabled: boolean`2400324004                    - `name: "bash"`2400524006                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2400724008                      Permission policy for tool execution.2400924010                      - `BetaManagedAgentsAlwaysAllowPolicy object`2401124012                        Tool calls are automatically approved without user confirmation.2401324014                        - `type: "always_allow"`2401524016                      - `BetaManagedAgentsAlwaysAskPolicy object`2401724018                        Tool calls require user confirmation before execution.2401924020                        - `type: "always_ask"`2402124022                      - `BetaManagedAgentsAutoPolicy object`2402324024                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2402524026                        - `type: "auto"`2402724028                  - `BetaManagedAgentsEditToolConfig object`2402924030                    Configuration for the edit tool.2403124032                    - `type: "edit"`2403324034                    - `enabled: boolean`2403524036                    - `name: "edit"`2403724038                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2403924040                      Permission policy for tool execution.2404124042                      - `BetaManagedAgentsAlwaysAllowPolicy object`2404324044                        Tool calls are automatically approved without user confirmation.2404524046                      - `BetaManagedAgentsAlwaysAskPolicy object`2404724048                        Tool calls require user confirmation before execution.2404924050                      - `BetaManagedAgentsAutoPolicy object`2405124052                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2405324054                  - `BetaManagedAgentsReadToolConfig object`2405524056                    Configuration for the read tool.2405724058                    - `type: "read"`2405924060                    - `enabled: boolean`2406124062                    - `name: "read"`2406324064                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2406524066                      Permission policy for tool execution.2406724068                      - `BetaManagedAgentsAlwaysAllowPolicy object`2406924070                        Tool calls are automatically approved without user confirmation.2407124072                      - `BetaManagedAgentsAlwaysAskPolicy object`2407324074                        Tool calls require user confirmation before execution.2407524076                      - `BetaManagedAgentsAutoPolicy object`2407724078                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2407924080                  - `BetaManagedAgentsWriteToolConfig object`2408124082                    Configuration for the write tool.2408324084                    - `type: "write"`2408524086                    - `enabled: boolean`2408724088                    - `name: "write"`2408924090                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2409124092                      Permission policy for tool execution.2409324094                      - `BetaManagedAgentsAlwaysAllowPolicy object`2409524096                        Tool calls are automatically approved without user confirmation.2409724098                      - `BetaManagedAgentsAlwaysAskPolicy object`2409924100                        Tool calls require user confirmation before execution.2410124102                      - `BetaManagedAgentsAutoPolicy object`2410324104                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2410524106                  - `BetaManagedAgentsGlobToolConfig object`2410724108                    Configuration for the glob tool.2410924110                    - `type: "glob"`2411124112                    - `enabled: boolean`2411324114                    - `name: "glob"`2411524116                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2411724118                      Permission policy for tool execution.2411924120                      - `BetaManagedAgentsAlwaysAllowPolicy object`2412124122                        Tool calls are automatically approved without user confirmation.2412324124                      - `BetaManagedAgentsAlwaysAskPolicy object`2412524126                        Tool calls require user confirmation before execution.2412724128                      - `BetaManagedAgentsAutoPolicy object`2412924130                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2413124132                  - `BetaManagedAgentsGrepToolConfig object`2413324134                    Configuration for the grep tool.2413524136                    - `type: "grep"`2413724138                    - `enabled: boolean`2413924140                    - `name: "grep"`2414124142                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2414324144                      Permission policy for tool execution.2414524146                      - `BetaManagedAgentsAlwaysAllowPolicy object`2414724148                        Tool calls are automatically approved without user confirmation.2414924150                      - `BetaManagedAgentsAlwaysAskPolicy object`2415124152                        Tool calls require user confirmation before execution.2415324154                      - `BetaManagedAgentsAutoPolicy object`2415524156                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2415724158                  - `BetaManagedAgentsWebFetchToolConfig object`2415924160                    Configuration for the web_fetch tool.2416124162                    - `type: "web_fetch"`2416324164                    - `enabled: boolean`2416524166                    - `name: "web_fetch"`2416724168                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2416924170                      Permission policy for tool execution.2417124172                      - `BetaManagedAgentsAlwaysAllowPolicy object`2417324174                        Tool calls are automatically approved without user confirmation.2417524176                      - `BetaManagedAgentsAlwaysAskPolicy object`2417724178                        Tool calls require user confirmation before execution.2417924180                      - `BetaManagedAgentsAutoPolicy object`2418124182                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2418324184                    - `allowed_domains: optional array of string`2418524186                    - `blocked_domains: optional array of string`2418724188                    - `max_content_tokens: optional number or null`2418924190                      format: int322419124192                  - `BetaManagedAgentsWebSearchToolConfig object`2419324194                    Configuration for the web_search tool.2419524196                    - `type: "web_search"`2419724198                    - `enabled: boolean`2419924200                    - `name: "web_search"`2420124202                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2420324204                      Permission policy for tool execution.2420524206                      - `BetaManagedAgentsAlwaysAllowPolicy object`2420724208                        Tool calls are automatically approved without user confirmation.2420924210                      - `BetaManagedAgentsAlwaysAskPolicy object`2421124212                        Tool calls require user confirmation before execution.2421324214                      - `BetaManagedAgentsAutoPolicy object`2421524216                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2421724218                    - `allowed_domains: optional array of string`2421924220                    - `blocked_domains: optional array of string`2422124222                    - `user_location: optional BetaManagedAgentsUserLocation or null`2422324224                      Approximate user location for search result localization.2422524226                      - `type: "approximate"`2422724228                        Location precision. Only "approximate" is supported.2422924230                      - `city: optional string or null`2423124232                        City name.2423324234                        minLength: 1, maxLength: 2552423524236                      - `country: optional string or null`2423724238                        Two-letter ISO 3166-1 country code, uppercase.2423924240                      - `region: optional string or null`2424124242                        Region or state name.2424324244                        minLength: 1, maxLength: 2552424524246                      - `timezone: optional string or null`2424724248                        IANA timezone identifier, e.g. "America/Los_Angeles".2424924250                        minLength: 1, maxLength: 2552425124252                - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`2425324254                  Resolved default configuration for agent tools.2425524256                  - `enabled: boolean`2425724258                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2425924260                    Permission policy for tool execution.2426124262                    - `BetaManagedAgentsAlwaysAllowPolicy object`2426324264                      Tool calls are automatically approved without user confirmation.2426524266                    - `BetaManagedAgentsAlwaysAskPolicy object`2426724268                      Tool calls require user confirmation before execution.2426924270                    - `BetaManagedAgentsAutoPolicy object`2427124272                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2427324274              - `BetaManagedAgentsMCPToolset object`2427524276                - `type: "mcp_toolset"`2427724278                - `configs: array of BetaManagedAgentsMCPToolConfig`2427924280                  - `enabled: boolean`2428124282                  - `name: string`2428324284                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2428524286                    Permission policy for tool execution.2428724288                    - `BetaManagedAgentsAlwaysAllowPolicy object`2428924290                      Tool calls are automatically approved without user confirmation.2429124292                    - `BetaManagedAgentsAlwaysAskPolicy object`2429324294                      Tool calls require user confirmation before execution.2429524296                    - `BetaManagedAgentsAutoPolicy object`2429724298                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2429924300                - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`2430124302                  Resolved default configuration for all tools from an MCP server.2430324304                  - `enabled: boolean`2430524306                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2430724308                    Permission policy for tool execution.2430924310                    - `BetaManagedAgentsAlwaysAllowPolicy object`2431124312                      Tool calls are automatically approved without user confirmation.2431324314                    - `BetaManagedAgentsAlwaysAskPolicy object`2431524316                      Tool calls require user confirmation before execution.2431724318                    - `BetaManagedAgentsAutoPolicy object`2431924320                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2432124322                - `mcp_server_name: string`2432324324              - `BetaManagedAgentsCustomTool object`2432524326                A custom tool as returned in API responses.2432724328                - `type: "custom"`2432924330                - `description: string`2433124332                - `input_schema: BetaManagedAgentsCustomToolInputSchema`2433324334                  JSON Schema for custom tool input parameters.2433524336                  - `type: "object"`2433724338                  - `properties: optional map[unknown] or null`2433924340                  - `required: optional array of string or null`2434124342                - `name: string`2434324344            - `version: number`2434524346              format: int322434724348          - `BetaManagedAgentsAdvisor object`2434924350            Platform advisor roster entry: a model the session's primary thread may consult mid-turn.2435124352            - `type: "advisor"`2435324354            - `model: string`2435524356              The advisor model id.2435724358      - `name: string`2435924360      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`2436124362        - `BetaManagedAgentsAnthropicSkill object`2436324364          A resolved Anthropic-managed skill.2436524366        - `BetaManagedAgentsCustomSkill object`2436724368          A resolved user-created custom skill.2436924370      - `system: string or null`2437124372      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`2437324374        - `BetaManagedAgentsAgentToolset20260401 object`2437524376        - `BetaManagedAgentsMCPToolset object`2437724378        - `BetaManagedAgentsCustomTool object`2437924380          A custom tool as returned in API responses.2438124382      - `version: number`2438324384        format: int322438524386    - `budget: optional BetaManagedAgentsBudgetLimit or null`2438724388      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.2438924390      - `type: "limit"`2439124392      - `max_list_cost: BetaMonetaryAmount`2439324394        A monetary amount in a specific currency.2439524396        - `amount: string`2439724398          Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.2439924400        - `currency: BetaCurrency`2440124402          Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.2440324404    - `metadata: optional map[string]`2440524406      The session's full metadata bag after the update. Present when the update set non-empty metadata; absent when metadata was unchanged or cleared to empty.2440724408    - `title: optional string or null`2440924410      The session's new title. Present only when the update changed it.2441124412  - `BetaManagedAgentsSystemMessageEvent object`2441324414    A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.2441524416    - `type: "system.message"`2441724418    - `id: string`2441924420      Unique identifier for this event.2442124422    - `content: array of BetaManagedAgentsSystemContentBlock`2442324424      System content blocks. Text-only.2442524426      - `type: "text"`2442724428      - `text: string`2442924430        The text content.2443124432        minLength: 12443324434    - `processed_at: optional string or null`2443524436      A timestamp in RFC 3339 format2443724438      format: date-time2443924440  - `BetaManagedAgentsSessionUsageEvent object`2444124442    Periodic snapshot of the session's cumulative usage and tracked list cost.2444324444    - `type: "session.usage"`2444524446    - `id: string`2444724448      Unique identifier for this event.2444924450    - `processed_at: string`2445124452      A timestamp in RFC 3339 format2445324454      format: date-time2445524456    - `usage: BetaManagedAgentsSessionUsageSnapshot`2445724458      Point-in-time snapshot of a session's cumulative usage.2445924460      - `active_seconds: optional number`2446124462        Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once. This is the duration the session's runtime cost is priced on.2446324464        format: double2446524466      - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`2446724468        Prompt-cache creation token usage broken down by cache lifetime.2446924470        - `ephemeral_1h_input_tokens: optional number`2447124472          Tokens used to create 1-hour ephemeral cache entries.2447324474          format: int322447524476        - `ephemeral_5m_input_tokens: optional number`2447724478          Tokens used to create 5-minute ephemeral cache entries.2447924480          format: int322448124482      - `cache_read_input_tokens: optional number`2448324484        Total tokens read from prompt cache.2448524486        format: int322448724488      - `input_tokens: optional number`2448924490        Total input tokens consumed across all turns.2449124492        format: int322449324494      - `list_cost: optional BetaMonetaryAmount`2449524496        A monetary amount in a specific currency.2449724498      - `output_tokens: optional number`2449924500        Total output tokens generated across all turns.2450124502        format: int322450324504      - `server_tool_use: optional BetaManagedAgentsServerToolUsage`2450524506        Cumulative count of server-executed tool invocations, broken down by tool.2450724508        - `web_fetch_requests: optional number`2450924510          Number of server-executed web fetch requests.2451124512          format: int322451324514        - `web_search_requests: optional number`2451524516          Number of server-executed web search requests.2451724518          format: int322451924520    - `budget: optional BetaManagedAgentsBudgetLimit or null`2452124522      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.2452324524- `next_page: optional string or null`2452524526  Opaque cursor for the next page. Null when no more results.2452724528#### Example2452924530```bash24531curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID/events \24532    -H 'anthropic-version: 2023-06-01' \24533    -H 'anthropic-beta: managed-agents-2026-04-01' \24534    -H "X-Api-Key: $ANTHROPIC_API_KEY"24535```2453624537##### Response (200)2453824539```json24540{24541  "data": [24542    {24543      "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",24544      "content": [24545        {24546          "text": "Where is my order #1234?",24547          "type": "text"24548        }24549      ],24550      "type": "user.message",24551      "processed_at": "2026-03-15T10:00:00Z"24552    }24553  ],24554  "next_page": "next_page"24555}24556```2455724558### Stream Session Thread Events2455924560**GET** `/v1/sessions/{session_id}/threads/{thread_id}/stream`2456124562Stream Session Thread Events2456324564#### Path parameters2456524566- `session_id: string`2456724568- `thread_id: string`2456924570#### Query parameters2457124572- `event_deltas: optional array of BetaManagedAgentsDeltaType`2457324574  When set, this connection also receives streaming deltas (`event_start`, `event_delta`) while an event is being produced, before the event itself arrives. Deltas are best-effort; when the final event is produced it carries the complete content. A model request that ends early (an error or interrupt) produces no final event — its terminal `span.model_request_end` closes the preview. Accepts one or more event types to preview and may be repeated: `agent.message` streams `content_delta` fragments; `agent.thinking` is start-only — a signal that the agent has begun extended thinking, concluded by the `agent.thinking` event itself. Only previews of the requested event types are sent.2457524576  - `"agent.message"`2457724578  - `"agent.thinking"`2457924580#### Headers2458124582- `"anthropic-beta": optional array of AnthropicBeta`2458324584  Optional header to specify the beta version(s) you want to use.2458524586  - `string`2458724588  - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`2458924590    - `"message-batches-2024-09-24"`2459124592    - `"prompt-caching-2024-07-31"`2459324594    - `"computer-use-2024-10-22"`2459524596    - `"computer-use-2025-01-24"`2459724598    - `"pdfs-2024-09-25"`2459924600    - `"token-counting-2024-11-01"`2460124602    - `"token-efficient-tools-2025-02-19"`2460324604    - `"output-128k-2025-02-19"`2460524606    - `"files-api-2025-04-14"`2460724608    - `"mcp-client-2025-04-04"`2460924610    - `"mcp-client-2025-11-20"`2461124612    - `"dev-full-thinking-2025-05-14"`2461324614    - `"interleaved-thinking-2025-05-14"`2461524616    - `"code-execution-2025-05-22"`2461724618    - `"extended-cache-ttl-2025-04-11"`2461924620    - `"context-1m-2025-08-07"`2462124622    - `"context-management-2025-06-27"`2462324624    - `"model-context-window-exceeded-2025-08-26"`2462524626    - `"skills-2025-10-02"`2462724628    - `"fast-mode-2026-02-01"`2462924630    - `"output-300k-2026-03-24"`2463124632    - `"user-profiles-2026-03-24"`2463324634    - `"user-profiles-2026-08-18"`2463524636    - `"user-profiles-2026-09-04"`2463724638    - `"advisor-tool-2026-03-01"`2463924640    - `"managed-agents-2026-04-01"`2464124642    - `"cache-diagnosis-2026-04-07"`2464324644    - `"dreaming-2026-04-21"`2464524646    - `"thinking-token-count-2026-05-13"`2464724648    - `"server-side-fallback-2026-06-01"`2464924650    - `"server-side-fallback-2026-07-01"`2465124652    - `"fallback-credit-2026-06-01"`2465324654    - `"fallback-credit-2026-07-01"`2465524656    - `"agent-memory-2026-07-22"`2465724658    - `"mid-conversation-tool-changes-2026-07-01"`2465924660    - `"compact-2026-01-12"`2466124662    - `"computer-use-2025-11-24"`2466324664    - `"mcp-tunnels-2026-06-22"`2466524666    - `"structured-outputs-2025-11-13"`2466724668    - `"task-budgets-2026-03-13"`2466924670    - `"thinking-display-updates-2026-08-18"`2467124672    - `"ce-user-management-2026-07-13"`2467324674    - `"mid-conversation-output-config-2026-07-01"`2467524676    - `"thinking-binding-controls-2026-08-01"`2467724678    - `"mid-conversation-system-clear-at-2026-08-21"`2467924680    - `"compact-2026-09-04"`2468124682- `"anthropic-workspace-id": optional string`2468324684#### Returns2468524686- `BetaManagedAgentsStreamSessionThreadEvents = BetaManagedAgentsUserMessageEvent or BetaManagedAgentsUserInterruptEvent or BetaManagedAgentsUserToolConfirmationEvent or 34 more`2468724688  Server-sent event in a single thread's stream.2468924690  - `BetaManagedAgentsUserMessageEvent object`2469124692    A user message event in the session conversation.2469324694    - `type: "user.message"`2469524696    - `id: string`2469724698      Unique identifier for this event.2469924700    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`2470124702      Array of content blocks comprising the user message.2470324704      - `BetaManagedAgentsTextBlock object`2470524706        Regular text content.2470724708        - `type: "text"`2470924710        - `text: string`2471124712          The text content.2471324714          minLength: 12471524716      - `BetaManagedAgentsImageBlock object`2471724718        Image content specified directly as base64 data or as a reference via a URL.2471924720        - `type: "image"`2472124722        - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`2472324724          Union type for image source variants.2472524726          - `BetaManagedAgentsBase64ImageSource object`2472724728            Base64-encoded image data.2472924730            - `type: "base64"`2473124732            - `data: string`2473324734              Base64-encoded image data.2473524736              minLength: 12473724738            - `media_type: string`2473924740              MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").2474124742              minLength: 12474324744          - `BetaManagedAgentsURLImageSource object`2474524746            Image referenced by URL.2474724748            - `type: "url"`2474924750            - `url: string`2475124752              URL of the image to fetch.2475324754              minLength: 12475524756          - `BetaManagedAgentsFileImageSource object`2475724758            Image referenced by file ID.2475924760            - `type: "file"`2476124762            - `file_id: string`2476324764              ID of a previously uploaded file.2476524766              minLength: 12476724768      - `BetaManagedAgentsDocumentBlock object`2476924770        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2477124772        - `type: "document"`2477324774        - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`2477524776          Union type for document source variants.2477724778          - `BetaManagedAgentsBase64DocumentSource object`2477924780            Base64-encoded document data.2478124782            - `type: "base64"`2478324784            - `data: string`2478524786              Base64-encoded document data.2478724788              minLength: 12478924790            - `media_type: string`2479124792              MIME type of the document (e.g., "application/pdf").2479324794              minLength: 12479524796          - `BetaManagedAgentsPlainTextDocumentSource object`2479724798            Plain text document content.2479924800            - `type: "text"`2480124802            - `data: string`2480324804              The plain text content.2480524806              minLength: 12480724808            - `media_type: "text/plain"`2480924810              MIME type of the text content. Must be "text/plain".2481124812          - `BetaManagedAgentsURLDocumentSource object`2481324814            Document referenced by URL.2481524816            - `type: "url"`2481724818            - `url: string`2481924820              URL of the document to fetch.2482124822              minLength: 12482324824          - `BetaManagedAgentsFileDocumentSource object`2482524826            Document referenced by file ID.2482724828            - `type: "file"`2482924830            - `file_id: string`2483124832              ID of a previously uploaded file.2483324834              minLength: 12483524836        - `context: optional string or null`2483724838          Additional context about the document for the model.2483924840        - `title: optional string or null`2484124842          The title of the document.2484324844      - `BetaManagedAgentsRedactedBlock object`2484524846        Placeholder for content withheld by Anthropic model policy.2484724848        - `type: "redacted"`2484924850    - `processed_at: optional string or null`2485124852      A timestamp in RFC 3339 format2485324854      format: date-time2485524856  - `BetaManagedAgentsUserInterruptEvent object`2485724858    An interrupt event that pauses agent execution and returns control to the user.2485924860    - `type: "user.interrupt"`2486124862    - `id: string`2486324864      Unique identifier for this event.2486524866    - `processed_at: optional string or null`2486724868      A timestamp in RFC 3339 format2486924870      format: date-time2487124872    - `session_thread_id: optional string or null`2487324874      If absent, interrupts every non-archived thread in a multiagent session (or the primary alone in a single-agent session). If present, interrupts only the named thread.2487524876  - `BetaManagedAgentsUserToolConfirmationEvent object`2487724878    A tool confirmation event that approves or denies a pending tool execution.2487924880    - `type: "user.tool_confirmation"`2488124882    - `id: string`2488324884      Unique identifier for this event.2488524886    - `result: "allow" or "deny"`2488724888      UserToolConfirmationResult enum2488924890      - `"allow"`2489124892      - `"deny"`2489324894    - `tool_use_id: string`2489524896      The id of the `agent.tool_use` or `agent.mcp_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.2489724898    - `deny_message: optional string or null`2489924900      Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.2490124902      maxLength: 100002490324904    - `processed_at: optional string or null`2490524906      A timestamp in RFC 3339 format2490724908      format: date-time2490924910    - `session_thread_id: optional string or null`2491124912      Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.2491324914  - `BetaManagedAgentsUserCustomToolResultEvent object`2491524916    Event sent by the client providing the result of a custom tool execution.2491724918    - `type: "user.custom_tool_result"`2491924920    - `id: string`2492124922      Unique identifier for this event.2492324924    - `custom_tool_use_id: string`2492524926      The id of the `agent.custom_tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.2492724928    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2492924930      The result content returned by the tool.2493124932      - `BetaManagedAgentsTextBlock object`2493324934        Regular text content.2493524936      - `BetaManagedAgentsImageBlock object`2493724938        Image content specified directly as base64 data or as a reference via a URL.2493924940      - `BetaManagedAgentsDocumentBlock object`2494124942        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2494324944      - `BetaManagedAgentsSearchResultBlock object`2494524946        A block containing a web search result.2494724948        - `type: "search_result"`2494924950        - `citations: BetaManagedAgentsSearchResultCitations`2495124952          Citation settings for a search result.2495324954          - `enabled: boolean`2495524956            Whether citations are enabled for this search result.2495724958        - `content: array of BetaManagedAgentsSearchResultContent`2495924960          Array of text content blocks from the search result.2496124962          - `type: "text"`2496324964          - `text: string`2496524966            The text content.2496724968            minLength: 12496924970        - `source: string`2497124972          The URL source of the search result.2497324974          minLength: 12497524976        - `title: string`2497724978          The title of the search result.2497924980          minLength: 12498124982    - `is_error: optional boolean or null`2498324984      Whether the tool execution resulted in an error.2498524986    - `processed_at: optional string or null`2498724988      A timestamp in RFC 3339 format2498924990      format: date-time2499124992    - `session_thread_id: optional string or null`2499324994      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.2499524996  - `BetaManagedAgentsAgentCustomToolUseEvent object`2499724998    Event emitted when the agent calls a custom tool. The session goes idle until the client sends a `user.custom_tool_result` event with the result.2499925000    - `type: "agent.custom_tool_use"`2500125002    - `id: string`2500325004      Unique identifier for this event.2500525006    - `input: map[unknown]`2500725008      Input parameters for the tool call.2500925010    - `name: string`2501125012      Name of the custom tool being called.2501325014    - `processed_at: string`2501525016      A timestamp in RFC 3339 format2501725018      format: date-time2501925020    - `session_thread_id: optional string or null`2502125022      When set, this event was cross-posted from a subagent's thread to surface its custom tool use on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.custom_tool_result` by `custom_tool_use_id`, so clients do not send it back.2502325024  - `BetaManagedAgentsAgentMessageEvent object`2502525026    An agent response event in the session conversation.2502725028    - `type: "agent.message"`2502925030    - `id: string`2503125032      Unique identifier for this event.2503325034    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`2503525036      Array of text blocks comprising the agent response.2503725038      - `BetaManagedAgentsTextBlock object`2503925040        Regular text content.2504125042      - `BetaManagedAgentsRedactedBlock object`2504325044        Placeholder for content withheld by Anthropic model policy.2504525046    - `processed_at: string`2504725048      A timestamp in RFC 3339 format2504925050      format: date-time2505125052  - `BetaManagedAgentsAgentThinkingEvent object`2505325054    Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.2505525056    - `type: "agent.thinking"`2505725058    - `id: string`2505925060      Unique identifier for this event.2506125062    - `processed_at: string`2506325064      A timestamp in RFC 3339 format2506525066      format: date-time2506725068  - `BetaManagedAgentsAgentMCPToolUseEvent object`2506925070    Event emitted when the agent invokes a tool provided by an MCP server.2507125072    - `type: "agent.mcp_tool_use"`2507325074    - `id: string`2507525076      Unique identifier for this event.2507725078    - `input: map[unknown]`2507925080      Input parameters for the tool call.2508125082    - `mcp_server_name: string`2508325084      Name of the MCP server providing the tool.2508525086    - `name: string`2508725088      Name of the MCP tool being used.2508925090    - `processed_at: string`2509125092      A timestamp in RFC 3339 format2509325094      format: date-time2509525096    - `evaluated_permission: optional "allow" or "ask" or "deny"`2509725098      AgentEvaluatedPermission enum2509925100      - `"allow"`2510125102      - `"ask"`2510325104      - `"deny"`2510525106    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`2510725108      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.2510925110      - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`2511125112        The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".2511325114        - `type: "always_allow"`2511525116      - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`2511725118        The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".2511925120        - `type: "always_ask"`2512125122      - `BetaManagedAgentsAgentToolEvaluationAuto object`2512325124        The resolved permission_policy was auto: the server judged this invocation individually.2512525126        - `type: "auto"`2512725128        - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`2512925130          The server's per-invocation judgement under the auto permission policy. Its type always equals the event's top-level evaluated_permission. Open union: clients must tolerate unknown variants.2513125132          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`2513325134            The server judged the invocation safe to execute without client approval.2513525136            - `type: "allow"`2513725138          - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`2513925140            The server reached no judgement; the invocation is held for client approval.2514125142            - `type: "ask"`2514325144            - `reason_code: string`2514525146              The judgement's grounds in registry-bound terms, for client branching and audit rather than end-user display. Open registry; currently "indeterminate" (no judgement was reached). Clients must tolerate values outside this set.2514725148              maxLength: 642514925150          - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`2515125152            The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.2515325154            - `type: "deny"`2515525156            - `reason_code: string`2515725158              The judgement's grounds in registry-bound terms. Open registry; currently "high_risk" (judged high-risk; the call does not run). Clients must tolerate values outside this set.2515925160              maxLength: 642516125162    - `session_thread_id: optional string or null`2516325164      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` by `tool_use_id`, so clients do not send it back.2516525166  - `BetaManagedAgentsAgentMCPToolResultEvent object`2516725168    Event representing the result of an MCP tool execution.2516925170    - `type: "agent.mcp_tool_result"`2517125172    - `id: string`2517325174      Unique identifier for this event.2517525176    - `mcp_tool_use_id: string`2517725178      The id of the `agent.mcp_tool_use` event this result corresponds to.2517925180    - `processed_at: string`2518125182      A timestamp in RFC 3339 format2518325184      format: date-time2518525186    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2518725188      The result content returned by the tool.2518925190      - `BetaManagedAgentsTextBlock object`2519125192        Regular text content.2519325194      - `BetaManagedAgentsImageBlock object`2519525196        Image content specified directly as base64 data or as a reference via a URL.2519725198      - `BetaManagedAgentsDocumentBlock object`2519925200        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2520125202      - `BetaManagedAgentsSearchResultBlock object`2520325204        A block containing a web search result.2520525206    - `is_error: optional boolean or null`2520725208      Whether the tool execution resulted in an error.2520925210  - `BetaManagedAgentsAgentToolUseEvent object`2521125212    Event emitted when the agent invokes a built-in agent tool.2521325214    - `type: "agent.tool_use"`2521525216    - `id: string`2521725218      Unique identifier for this event.2521925220    - `input: map[unknown]`2522125222      Input parameters for the tool call.2522325224    - `name: string`2522525226      Name of the agent tool being used.2522725228    - `processed_at: string`2522925230      A timestamp in RFC 3339 format2523125232      format: date-time2523325234    - `evaluated_permission: optional "allow" or "ask" or "deny"`2523525236      AgentEvaluatedPermission enum2523725238      - `"allow"`2523925240      - `"ask"`2524125242      - `"deny"`2524325244    - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`2524525246      Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.2524725248    - `session_thread_id: optional string or null`2524925250      When set, this event was cross-posted from a subagent's thread to surface its permission request on the primary thread's stream. Empty on the thread's own events. Informational only: the server routes the matching `user.tool_confirmation` or `user.tool_result` by `tool_use_id`, so clients do not send it back.2525125252  - `BetaManagedAgentsAgentToolResultEvent object`2525325254    Event representing the result of an agent tool execution.2525525256    - `type: "agent.tool_result"`2525725258    - `id: string`2525925260      Unique identifier for this event.2526125262    - `processed_at: string`2526325264      A timestamp in RFC 3339 format2526525266      format: date-time2526725268    - `tool_use_id: string`2526925270      The id of the `agent.tool_use` event this result corresponds to.2527125272    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2527325274      The result content returned by the tool.2527525276      - `BetaManagedAgentsTextBlock object`2527725278        Regular text content.2527925280      - `BetaManagedAgentsImageBlock object`2528125282        Image content specified directly as base64 data or as a reference via a URL.2528325284      - `BetaManagedAgentsDocumentBlock object`2528525286        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2528725288      - `BetaManagedAgentsSearchResultBlock object`2528925290        A block containing a web search result.2529125292    - `is_error: optional boolean or null`2529325294      Whether the tool execution resulted in an error.2529525296  - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`2529725298    Delivery event written to the target thread's input stream when an agent-to-agent message arrives.2529925300    - `type: "agent.thread_message_received"`2530125302    - `id: string`2530325304      Unique identifier for this event.2530525306    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`2530725308      Message content blocks.2530925310      - `BetaManagedAgentsTextBlock object`2531125312        Regular text content.2531325314      - `BetaManagedAgentsImageBlock object`2531525316        Image content specified directly as base64 data or as a reference via a URL.2531725318      - `BetaManagedAgentsDocumentBlock object`2531925320        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2532125322      - `BetaManagedAgentsRedactedBlock object`2532325324        Placeholder for content withheld by Anthropic model policy.2532525326    - `from_session_thread_id: string`2532725328      Public `sthr_` ID of the thread that sent the message.2532925330    - `processed_at: string`2533125332      A timestamp in RFC 3339 format2533325334      format: date-time2533525336    - `from_agent_name: optional string or null`2533725338      Name of the callable agent this message came from. Absent when received from the primary agent.2533925340  - `BetaManagedAgentsAgentThreadMessageSentEvent object`2534125342    Observability event emitted to the sender's output stream when an agent-to-agent message is sent.2534325344    - `type: "agent.thread_message_sent"`2534525346    - `id: string`2534725348      Unique identifier for this event.2534925350    - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`2535125352      Message content blocks.2535325354      - `BetaManagedAgentsTextBlock object`2535525356        Regular text content.2535725358      - `BetaManagedAgentsImageBlock object`2535925360        Image content specified directly as base64 data or as a reference via a URL.2536125362      - `BetaManagedAgentsDocumentBlock object`2536325364        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2536525366      - `BetaManagedAgentsRedactedBlock object`2536725368        Placeholder for content withheld by Anthropic model policy.2536925370    - `processed_at: string`2537125372      A timestamp in RFC 3339 format2537325374      format: date-time2537525376    - `to_session_thread_id: string`2537725378      Public `sthr_` ID of the thread the message was sent to.2537925380    - `to_agent_name: optional string or null`2538125382      Name of the callable agent this message was sent to. Absent when sent to the primary agent.2538325384  - `BetaManagedAgentsAgentThreadContextCompactedEvent object`2538525386    Indicates that context compaction (summarization) occurred during the session.2538725388    - `type: "agent.thread_context_compacted"`2538925390    - `id: string`2539125392      Unique identifier for this event.2539325394    - `processed_at: string`2539525396      A timestamp in RFC 3339 format2539725398      format: date-time2539925400  - `BetaManagedAgentsSessionErrorEvent object`2540125402    An error event indicating a problem occurred during session execution.2540325404    - `type: "session.error"`2540525406    - `id: string`2540725408      Unique identifier for this event.2540925410    - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`2541125412      - `BetaManagedAgentsUnknownError object`2541325414        An unknown or unexpected error occurred during session execution. A fallback variant; clients that don't recognize a new error code can match on `retry_status` and `message` alone.2541525416        - `type: "unknown_error"`2541725418        - `message: string`2541925420          Human-readable error description.2542125422        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2542325424          What the client should do next in response to this error.2542525426          - `BetaManagedAgentsRetryStatusRetrying object`2542725428            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2542925430            - `type: "retrying"`2543125432          - `BetaManagedAgentsRetryStatusExhausted object`2543325434            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2543525436            - `type: "exhausted"`2543725438          - `BetaManagedAgentsRetryStatusTerminal object`2543925440            The session encountered a terminal error and will transition to `terminated` state.2544125442            - `type: "terminal"`2544325444      - `BetaManagedAgentsModelOverloadedError object`2544525446        The model is currently overloaded. Emitted after automatic retries are exhausted.2544725448        - `type: "model_overloaded_error"`2544925450        - `message: string`2545125452          Human-readable error description.2545325454        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2545525456          What the client should do next in response to this error.2545725458          - `BetaManagedAgentsRetryStatusRetrying object`2545925460            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2546125462          - `BetaManagedAgentsRetryStatusExhausted object`2546325464            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2546525466          - `BetaManagedAgentsRetryStatusTerminal object`2546725468            The session encountered a terminal error and will transition to `terminated` state.2546925470      - `BetaManagedAgentsModelRateLimitedError object`2547125472        The model request was rate-limited.2547325474        - `type: "model_rate_limited_error"`2547525476        - `message: string`2547725478          Human-readable error description.2547925480        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2548125482          What the client should do next in response to this error.2548325484          - `BetaManagedAgentsRetryStatusRetrying object`2548525486            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2548725488          - `BetaManagedAgentsRetryStatusExhausted object`2548925490            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2549125492          - `BetaManagedAgentsRetryStatusTerminal object`2549325494            The session encountered a terminal error and will transition to `terminated` state.2549525496      - `BetaManagedAgentsModelRequestFailedError object`2549725498        A model request failed for a reason other than overload or rate-limiting.2549925500        - `type: "model_request_failed_error"`2550125502        - `message: string`2550325504          Human-readable error description.2550525506        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2550725508          What the client should do next in response to this error.2550925510          - `BetaManagedAgentsRetryStatusRetrying object`2551125512            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2551325514          - `BetaManagedAgentsRetryStatusExhausted object`2551525516            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2551725518          - `BetaManagedAgentsRetryStatusTerminal object`2551925520            The session encountered a terminal error and will transition to `terminated` state.2552125522      - `BetaManagedAgentsMCPConnectionFailedError object`2552325524        Failed to connect to an MCP server.2552525526        - `type: "mcp_connection_failed_error"`2552725528        - `mcp_server_name: string`2552925530          Name of the MCP server that failed to connect.2553125532        - `message: string`2553325534          Human-readable error description.2553525536        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2553725538          What the client should do next in response to this error.2553925540          - `BetaManagedAgentsRetryStatusRetrying object`2554125542            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2554325544          - `BetaManagedAgentsRetryStatusExhausted object`2554525546            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2554725548          - `BetaManagedAgentsRetryStatusTerminal object`2554925550            The session encountered a terminal error and will transition to `terminated` state.2555125552      - `BetaManagedAgentsMCPAuthenticationFailedError object`2555325554        Authentication to an MCP server failed.2555525556        - `type: "mcp_authentication_failed_error"`2555725558        - `mcp_server_name: string`2555925560          Name of the MCP server that failed authentication.2556125562        - `message: string`2556325564          Human-readable error description.2556525566        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2556725568          What the client should do next in response to this error.2556925570          - `BetaManagedAgentsRetryStatusRetrying object`2557125572            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2557325574          - `BetaManagedAgentsRetryStatusExhausted object`2557525576            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2557725578          - `BetaManagedAgentsRetryStatusTerminal object`2557925580            The session encountered a terminal error and will transition to `terminated` state.2558125582      - `BetaManagedAgentsBillingError object`2558325584        The caller's organization or workspace cannot make model requests — out of credits or spend limit reached. Retrying with the same credentials will not succeed; the caller must resolve the billing state.2558525586        - `type: "billing_error"`2558725588        - `message: string`2558925590          Human-readable error description.2559125592        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2559325594          What the client should do next in response to this error.2559525596          - `BetaManagedAgentsRetryStatusRetrying object`2559725598            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2559925600          - `BetaManagedAgentsRetryStatusExhausted object`2560125602            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2560325604          - `BetaManagedAgentsRetryStatusTerminal object`2560525606            The session encountered a terminal error and will transition to `terminated` state.2560725608      - `BetaManagedAgentsCredentialHostUnreachableError object`2560925610        An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.2561125612        - `type: "credential_host_unreachable_error"`2561325614        - `credential_id: string`2561525616          ID of the affected credential.2561725618        - `message: string`2561925620          Human-readable error description.2562125622        - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`2562325624          What the client should do next in response to this error.2562525626          - `BetaManagedAgentsRetryStatusRetrying object`2562725628            The server is retrying automatically. Client should wait; the same error type may fire again as retrying, then once as exhausted when the retry budget runs out.2562925630          - `BetaManagedAgentsRetryStatusExhausted object`2563125632            This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.2563325634          - `BetaManagedAgentsRetryStatusTerminal object`2563525636            The session encountered a terminal error and will transition to `terminated` state.2563725638        - `vault_id: string`2563925640          ID of the vault containing the affected credential.2564125642    - `processed_at: string`2564325644      A timestamp in RFC 3339 format2564525646      format: date-time2564725648  - `BetaManagedAgentsSessionStatusRescheduledEvent object`2564925650    Indicates the session is recovering from an error state and is rescheduled for execution.2565125652    - `type: "session.status_rescheduled"`2565325654    - `id: string`2565525656      Unique identifier for this event.2565725658    - `processed_at: string`2565925660      A timestamp in RFC 3339 format2566125662      format: date-time2566325664  - `BetaManagedAgentsSessionStatusRunningEvent object`2566525666    Indicates the session is actively running and the agent is working.2566725668    - `type: "session.status_running"`2566925670    - `id: string`2567125672      Unique identifier for this event.2567325674    - `processed_at: string`2567525676      A timestamp in RFC 3339 format2567725678      format: date-time2567925680  - `BetaManagedAgentsSessionStatusIdleEvent object`2568125682    Indicates the agent has paused and is awaiting user input.2568325684    - `type: "session.status_idle"`2568525686    - `id: string`2568725688      Unique identifier for this event.2568925690    - `processed_at: string`2569125692      A timestamp in RFC 3339 format2569325694      format: date-time2569525696    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`2569725698      - `BetaManagedAgentsSessionEndTurn object`2569925700        The agent completed its turn naturally and is ready for the next user message.2570125702        - `type: "end_turn"`2570325704      - `BetaManagedAgentsSessionRequiresAction object`2570525706        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.2570725708        - `type: "requires_action"`2570925710        - `event_ids: array of string`2571125712          The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.2571325714      - `BetaManagedAgentsSessionRetriesExhausted object`2571525716        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.2571725718        - `type: "retries_exhausted"`2571925720      - `BetaManagedAgentsSessionBudgetReached object`2572125722        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.2572325724        - `type: "budget_reached"`2572525726  - `BetaManagedAgentsSessionStatusTerminatedEvent object`2572725728    Indicates the session has terminated, either due to an error or completion.2572925730    - `type: "session.status_terminated"`2573125732    - `id: string`2573325734      Unique identifier for this event.2573525736    - `processed_at: string`2573725738      A timestamp in RFC 3339 format2573925740      format: date-time2574125742  - `BetaManagedAgentsSessionThreadCreatedEvent object`2574325744    Emitted when a subagent is spawned as a new thread. Written to the parent thread's output stream so clients observing the session see child creation.2574525746    - `type: "session.thread_created"`2574725748    - `id: string`2574925750      Unique identifier for this event.2575125752    - `agent_name: string`2575325754      Name of the callable agent the thread runs.2575525756    - `processed_at: string`2575725758      A timestamp in RFC 3339 format2575925760      format: date-time2576125762    - `session_thread_id: string`2576325764      Public `sthr_` ID of the newly created thread.2576525766  - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`2576725768    Emitted when an outcome evaluation cycle begins.2576925770    - `type: "span.outcome_evaluation_start"`2577125772    - `id: string`2577325774      Unique identifier for this event.2577525776    - `iteration: number`2577725778      0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.2577925780      format: int322578125782    - `outcome_id: string`2578325784      The `outc_` ID of the outcome being evaluated.2578525786    - `processed_at: string`2578725788      A timestamp in RFC 3339 format2578925790      format: date-time2579125792  - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`2579325794    Emitted when an outcome evaluation cycle completes. Carries the verdict and aggregate token usage. A verdict of `needs_revision` means another evaluation cycle follows; `satisfied`, `max_iterations_reached`, `failed`, or `interrupted` are terminal — no further evaluation cycles follow.2579525796    - `type: "span.outcome_evaluation_end"`2579725798    - `id: string`2579925800      Unique identifier for this event.2580125802    - `explanation: string`2580325804      Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.2580525806    - `iteration: number`2580725808      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.2580925810      format: int322581125812    - `outcome_evaluation_start_id: string`2581325814      The id of the corresponding `span.outcome_evaluation_start` event.2581525816    - `outcome_id: string`2581725818      The `outc_` ID of the outcome being evaluated.2581925820    - `processed_at: string`2582125822      A timestamp in RFC 3339 format2582325824      format: date-time2582525826    - `result: string`2582725828      Evaluation verdict. 'satisfied': criteria met, session goes idle. 'needs_revision': criteria not met, another revision cycle follows. 'max_iterations_reached': evaluation budget exhausted with criteria still unmet — one final acknowledgment turn follows before the session goes idle, but no further evaluation runs. 'failed': grader determined the rubric does not apply to the deliverables. 'interrupted': user sent an interrupt while evaluation was in progress.2582925830    - `usage: BetaManagedAgentsSpanModelUsage`2583125832      Token usage for a single model request.2583325834      - `cache_creation_input_tokens: number`2583525836        Tokens used to create prompt cache in this request.2583725838        format: int322583925840      - `cache_read_input_tokens: number`2584125842        Tokens read from prompt cache in this request.2584325844        format: int322584525846      - `input_tokens: number`2584725848        Input tokens consumed by this request.2584925850        format: int322585125852      - `output_tokens: number`2585325854        Output tokens generated by this request.2585525856        format: int322585725858      - `speed: optional "standard" or "fast" or null`2585925860        Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.2586125862        - `"standard"`2586325864        - `"fast"`2586525866  - `BetaManagedAgentsSpanModelRequestStartEvent object`2586725868    Emitted when a model request is initiated by the agent.2586925870    - `type: "span.model_request_start"`2587125872    - `id: string`2587325874      Unique identifier for this event.2587525876    - `processed_at: string`2587725878      A timestamp in RFC 3339 format2587925880      format: date-time2588125882  - `BetaManagedAgentsSpanModelRequestEndEvent object`2588325884    Emitted when a model request completes.2588525886    - `type: "span.model_request_end"`2588725888    - `id: string`2588925890      Unique identifier for this event.2589125892    - `is_error: boolean or null`2589325894      Whether the model request resulted in an error.2589525896    - `model_request_start_id: string`2589725898      The id of the corresponding `span.model_request_start` event.2589925900    - `model_usage: BetaManagedAgentsSpanModelUsage`2590125902      Token usage for a single model request.2590325904    - `processed_at: string`2590525906      A timestamp in RFC 3339 format2590725908      format: date-time2590925910  - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`2591125912    Periodic heartbeat emitted while an outcome evaluation cycle is in progress. Distinguishes 'evaluation is actively running' from 'evaluation is stuck' between the corresponding `span.outcome_evaluation_start` and `span.outcome_evaluation_end` events.2591325914    - `type: "span.outcome_evaluation_ongoing"`2591525916    - `id: string`2591725918      Unique identifier for this event.2591925920    - `iteration: number`2592125922      0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.2592325924      format: int322592525926    - `outcome_id: string`2592725928      The `outc_` ID of the outcome being evaluated.2592925930    - `processed_at: string`2593125932      A timestamp in RFC 3339 format2593325934      format: date-time2593525936  - `BetaManagedAgentsUserDefineOutcomeEvent object`2593725938    Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.2593925940    - `type: "user.define_outcome"`2594125942    - `id: string`2594325944      Unique identifier for this event.2594525946    - `description: string`2594725948      What the agent should produce. Copied from the input event.2594925950    - `max_iterations: number or null`2595125952      Evaluate-then-revise cycles before giving up. Default 3, max 20.2595325954      format: int322595525956    - `outcome_id: string`2595725958      Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.2595925960    - `processed_at: string`2596125962      A timestamp in RFC 3339 format2596325964      format: date-time2596525966    - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`2596725968      Rubric for grading the quality of an outcome.2596925970      - `BetaManagedAgentsFileRubric object`2597125972        Rubric referenced by a file uploaded via the Files API.2597325974        - `type: "file"`2597525976        - `file_id: string`2597725978          ID of the rubric file.2597925980      - `BetaManagedAgentsTextRubric object`2598125982        Rubric content provided inline as text.2598325984        - `type: "text"`2598525986        - `content: string`2598725988          Rubric content. Plain text or markdown — the grader treats it as freeform text.2598925990  - `BetaManagedAgentsSessionDeletedEvent object`2599125992    Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.2599325994    - `type: "session.deleted"`2599525996    - `id: string`2599725998      Unique identifier for this event.2599926000    - `processed_at: string`2600126002      A timestamp in RFC 3339 format2600326004      format: date-time2600526006  - `BetaManagedAgentsSessionThreadStatusRunningEvent object`2600726008    A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2600926010    - `type: "session.thread_status_running"`2601126012    - `id: string`2601326014      Unique identifier for this event.2601526016    - `agent_name: string`2601726018      Name of the agent the thread runs.2601926020    - `processed_at: string`2602126022      A timestamp in RFC 3339 format2602326024      format: date-time2602526026    - `session_thread_id: string`2602726028      Public sthr_ ID of the thread that started running.2602926030  - `BetaManagedAgentsSessionThreadStatusIdleEvent object`2603126032    A session thread has yielded and is awaiting input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2603326034    - `type: "session.thread_status_idle"`2603526036    - `id: string`2603726038      Unique identifier for this event.2603926040    - `agent_name: string`2604126042      Name of the agent the thread runs.2604326044    - `processed_at: string`2604526046      A timestamp in RFC 3339 format2604726048      format: date-time2604926050    - `session_thread_id: string`2605126052      Public sthr_ ID of the thread that went idle.2605326054    - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`2605526056      - `BetaManagedAgentsSessionEndTurn object`2605726058        The agent completed its turn naturally and is ready for the next user message.2605926060      - `BetaManagedAgentsSessionRequiresAction object`2606126062        The agent is idle waiting on one or more blocking user-input events (tool confirmation, custom tool result, etc.). Resolving all of them transitions the session back to running.2606326064      - `BetaManagedAgentsSessionRetriesExhausted object`2606526066        The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.2606726068      - `BetaManagedAgentsSessionBudgetReached object`2606926070        The agent stopped because the session's tracked list cost reached its budget, or because its usage includes a model with no list price (which the budget cannot measure). Raise the budget to continue — or, if raising is rejected because a model has no list price, remove the budget.2607126072  - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`2607326074    A session thread has terminated and will accept no further input. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2607526076    - `type: "session.thread_status_terminated"`2607726078    - `id: string`2607926080      Unique identifier for this event.2608126082    - `agent_name: string`2608326084      Name of the agent the thread runs.2608526086    - `processed_at: string`2608726088      A timestamp in RFC 3339 format2608926090      format: date-time2609126092    - `session_thread_id: string`2609326094      Public sthr_ ID of the thread that terminated.2609526096  - `BetaManagedAgentsUserToolResultEvent object`2609726098    Event sent by the client providing the result of an agent-toolset tool execution. Only valid on `self_hosted` environments, where sandbox-routed tools are executed by the client rather than the server.2609926100    - `type: "user.tool_result"`2610126102    - `id: string`2610326104      Unique identifier for this event.2610526106    - `tool_use_id: string`2610726108      The id of the `agent.tool_use` event this result corresponds to, which can be found in the last `session.status_idle` [event's](https://platform.claude.com/docs/en/api/beta/sessions/events/list#beta_managed_agents_session_requires_action.event_ids) `stop_reason.event_ids` field.2610926110    - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`2611126112      The result content returned by the tool.2611326114      - `BetaManagedAgentsTextBlock object`2611526116        Regular text content.2611726118      - `BetaManagedAgentsImageBlock object`2611926120        Image content specified directly as base64 data or as a reference via a URL.2612126122      - `BetaManagedAgentsDocumentBlock object`2612326124        Document content, either specified directly as base64 data, as text, or as a reference via a URL.2612526126      - `BetaManagedAgentsSearchResultBlock object`2612726128        A block containing a web search result.2612926130    - `is_error: optional boolean or null`2613126132      Whether the tool execution resulted in an error.2613326134    - `processed_at: optional string or null`2613526136      A timestamp in RFC 3339 format2613726138      format: date-time2613926140    - `session_thread_id: optional string or null`2614126142      Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.2614326144  - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`2614526146    A session thread hit a transient error and is retrying automatically. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.2614726148    - `type: "session.thread_status_rescheduled"`2614926150    - `id: string`2615126152      Unique identifier for this event.2615326154    - `agent_name: string`2615526156      Name of the agent the thread runs.2615726158    - `processed_at: string`2615926160      A timestamp in RFC 3339 format2616126162      format: date-time2616326164    - `session_thread_id: string`2616526166      Public sthr_ ID of the thread that is retrying.2616726168  - `BetaManagedAgentsSessionUpdatedEvent object`2616926170    Emitted when an UpdateSession request changed at least one field. Carries only the fields that changed; absent fields were not part of the update. The new configuration applies from the next turn.2617126172    - `type: "session.updated"`2617326174    - `id: string`2617526176      Unique identifier for this event.2617726178    - `processed_at: string`2617926180      A timestamp in RFC 3339 format2618126182      format: date-time2618326184    - `agent: optional BetaManagedAgentsSessionAgent or null`2618526186      Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.2618726188      - `type: "agent"`2618926190      - `id: string`2619126192      - `description: string or null`2619326194      - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`2619526196        - `type: "url"`2619726198        - `name: string`2619926200        - `url: string`2620126202      - `model: BetaManagedAgentsModelConfig`2620326204        Model identifier and configuration.2620526206        - `id: BetaManagedAgentsModel`2620726208          The model that will power your agent.2620926210          See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2621126212          - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`2621326214            The model that will power your agent.2621526216            See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.2621726218            - `"claude-fable-5-1"`2621926220              Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows2622126222            - `"claude-sonnet-5"`2622326224              High-performance model for coding and agents2622526226            - `"claude-fable-5"`2622726228              Next generation of intelligence for the hardest knowledge work and coding problems2622926230            - `"claude-opus-5"`2623126232              Powerful intelligence for long-running agents and coding2623326234            - `"claude-opus-4-8"`2623526236              Powerful intelligence for long-running agents and coding2623726238            - `"claude-opus-4-7"`2623926240              Powerful intelligence for long-running agents and coding2624126242            - `"claude-opus-4-6"`2624326244              Powerful intelligence for long-running agents and coding2624526246            - `"claude-sonnet-4-6"`2624726248              Best combination of speed and intelligence2624926250            - `"claude-haiku-4-5"`2625126252              Fastest model with near-frontier intelligence2625326254            - `"claude-haiku-4-5-20251001"`2625526256              Fastest model with near-frontier intelligence2625726258            - `"claude-opus-4-5"`2625926260              Powerful intelligence for long-running agents and coding2626126262            - `"claude-opus-4-5-20251101"`2626326264              Powerful intelligence for long-running agents and coding2626526266            - `"claude-sonnet-4-5"`2626726268              High-performance model for agents and coding2626926270            - `"claude-sonnet-4-5-20250929"`2627126272              High-performance model for agents and coding2627326274          - `string`2627526276        - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`2627726278          How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.2627926280          - `BetaManagedAgentsEffortLow object`2628126282            Low effort. Favors latency over reasoning depth.2628326284            - `type: "low"`2628526286          - `BetaManagedAgentsEffortMedium object`2628726288            Medium effort. Balances latency and reasoning depth.2628926290            - `type: "medium"`2629126292          - `BetaManagedAgentsEffortHigh object`2629326294            High effort. Favors reasoning depth.2629526296            - `type: "high"`2629726298          - `BetaManagedAgentsEffortXhigh object`2629926300            Extra-high effort. Not all models accept this level.2630126302            - `type: "xhigh"`2630326304          - `BetaManagedAgentsEffortMax object`2630526306            Maximum effort. Favors reasoning depth over latency.2630726308            - `type: "max"`2630926310        - `inference_geo: optional string`2631126312          Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.2631326314        - `speed: optional "standard" or "fast"`2631526316          Inference speed mode. `fast` provides significantly faster output token generation at premium pricing. Not all models support `fast`; invalid combinations are rejected at create time.2631726318          - `"standard"`2631926320          - `"fast"`2632126322      - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`2632326324        Resolved coordinator topology with full agent definitions for each roster member.2632526326        - `type: "coordinator"`2632726328        - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`2632926330          Full `agent` definitions the coordinator may spawn as session threads.2633126332          - `BetaManagedAgentsSessionThreadAgent object`2633326334            Resolved `agent` definition for a single `session_thread`. Snapshot of the agent at thread creation time. The multiagent roster is not repeated here; read it from `Session.agent`.2633526336            - `type: "agent"`2633726338            - `id: string`2633926340            - `description: string or null`2634126342            - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`2634326344              - `type: "url"`2634526346              - `name: string`2634726348              - `url: string`2634926350            - `model: BetaManagedAgentsModelConfig`2635126352              Model identifier and configuration.2635326354            - `name: string`2635526356            - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`2635726358              - `BetaManagedAgentsAnthropicSkill object`2635926360                A resolved Anthropic-managed skill.2636126362                - `type: "anthropic"`2636326364                - `skill_id: string`2636526366                - `version: string`2636726368              - `BetaManagedAgentsCustomSkill object`2636926370                A resolved user-created custom skill.2637126372                - `type: "custom"`2637326374                - `skill_id: string`2637526376                - `version: string`2637726378            - `system: string or null`2637926380            - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`2638126382              - `BetaManagedAgentsAgentToolset20260401 object`2638326384                - `type: "agent_toolset_20260401"`2638526386                - `configs: array of BetaManagedAgentsAgentToolConfig`2638726388                  - `BetaManagedAgentsBashToolConfig object`2638926390                    Configuration for the bash tool.2639126392                    - `type: "bash"`2639326394                    - `enabled: boolean`2639526396                    - `name: "bash"`2639726398                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2639926400                      Permission policy for tool execution.2640126402                      - `BetaManagedAgentsAlwaysAllowPolicy object`2640326404                        Tool calls are automatically approved without user confirmation.2640526406                        - `type: "always_allow"`2640726408                      - `BetaManagedAgentsAlwaysAskPolicy object`2640926410                        Tool calls require user confirmation before execution.2641126412                        - `type: "always_ask"`2641326414                      - `BetaManagedAgentsAutoPolicy object`2641526416                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2641726418                        - `type: "auto"`2641926420                  - `BetaManagedAgentsEditToolConfig object`2642126422                    Configuration for the edit tool.2642326424                    - `type: "edit"`2642526426                    - `enabled: boolean`2642726428                    - `name: "edit"`2642926430                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2643126432                      Permission policy for tool execution.2643326434                      - `BetaManagedAgentsAlwaysAllowPolicy object`2643526436                        Tool calls are automatically approved without user confirmation.2643726438                      - `BetaManagedAgentsAlwaysAskPolicy object`2643926440                        Tool calls require user confirmation before execution.2644126442                      - `BetaManagedAgentsAutoPolicy object`2644326444                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2644526446                  - `BetaManagedAgentsReadToolConfig object`2644726448                    Configuration for the read tool.2644926450                    - `type: "read"`2645126452                    - `enabled: boolean`2645326454                    - `name: "read"`2645526456                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2645726458                      Permission policy for tool execution.2645926460                      - `BetaManagedAgentsAlwaysAllowPolicy object`2646126462                        Tool calls are automatically approved without user confirmation.2646326464                      - `BetaManagedAgentsAlwaysAskPolicy object`2646526466                        Tool calls require user confirmation before execution.2646726468                      - `BetaManagedAgentsAutoPolicy object`2646926470                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2647126472                  - `BetaManagedAgentsWriteToolConfig object`2647326474                    Configuration for the write tool.2647526476                    - `type: "write"`2647726478                    - `enabled: boolean`2647926480                    - `name: "write"`2648126482                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2648326484                      Permission policy for tool execution.2648526486                      - `BetaManagedAgentsAlwaysAllowPolicy object`2648726488                        Tool calls are automatically approved without user confirmation.2648926490                      - `BetaManagedAgentsAlwaysAskPolicy object`2649126492                        Tool calls require user confirmation before execution.2649326494                      - `BetaManagedAgentsAutoPolicy object`2649526496                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2649726498                  - `BetaManagedAgentsGlobToolConfig object`2649926500                    Configuration for the glob tool.2650126502                    - `type: "glob"`2650326504                    - `enabled: boolean`2650526506                    - `name: "glob"`2650726508                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2650926510                      Permission policy for tool execution.2651126512                      - `BetaManagedAgentsAlwaysAllowPolicy object`2651326514                        Tool calls are automatically approved without user confirmation.2651526516                      - `BetaManagedAgentsAlwaysAskPolicy object`2651726518                        Tool calls require user confirmation before execution.2651926520                      - `BetaManagedAgentsAutoPolicy object`2652126522                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2652326524                  - `BetaManagedAgentsGrepToolConfig object`2652526526                    Configuration for the grep tool.2652726528                    - `type: "grep"`2652926530                    - `enabled: boolean`2653126532                    - `name: "grep"`2653326534                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2653526536                      Permission policy for tool execution.2653726538                      - `BetaManagedAgentsAlwaysAllowPolicy object`2653926540                        Tool calls are automatically approved without user confirmation.2654126542                      - `BetaManagedAgentsAlwaysAskPolicy object`2654326544                        Tool calls require user confirmation before execution.2654526546                      - `BetaManagedAgentsAutoPolicy object`2654726548                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2654926550                  - `BetaManagedAgentsWebFetchToolConfig object`2655126552                    Configuration for the web_fetch tool.2655326554                    - `type: "web_fetch"`2655526556                    - `enabled: boolean`2655726558                    - `name: "web_fetch"`2655926560                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2656126562                      Permission policy for tool execution.2656326564                      - `BetaManagedAgentsAlwaysAllowPolicy object`2656526566                        Tool calls are automatically approved without user confirmation.2656726568                      - `BetaManagedAgentsAlwaysAskPolicy object`2656926570                        Tool calls require user confirmation before execution.2657126572                      - `BetaManagedAgentsAutoPolicy object`2657326574                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2657526576                    - `allowed_domains: optional array of string`2657726578                    - `blocked_domains: optional array of string`2657926580                    - `max_content_tokens: optional number or null`2658126582                      format: int322658326584                  - `BetaManagedAgentsWebSearchToolConfig object`2658526586                    Configuration for the web_search tool.2658726588                    - `type: "web_search"`2658926590                    - `enabled: boolean`2659126592                    - `name: "web_search"`2659326594                    - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2659526596                      Permission policy for tool execution.2659726598                      - `BetaManagedAgentsAlwaysAllowPolicy object`2659926600                        Tool calls are automatically approved without user confirmation.2660126602                      - `BetaManagedAgentsAlwaysAskPolicy object`2660326604                        Tool calls require user confirmation before execution.2660526606                      - `BetaManagedAgentsAutoPolicy object`2660726608                        The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2660926610                    - `allowed_domains: optional array of string`2661126612                    - `blocked_domains: optional array of string`2661326614                    - `user_location: optional BetaManagedAgentsUserLocation or null`2661526616                      Approximate user location for search result localization.2661726618                      - `type: "approximate"`2661926620                        Location precision. Only "approximate" is supported.2662126622                      - `city: optional string or null`2662326624                        City name.2662526626                        minLength: 1, maxLength: 2552662726628                      - `country: optional string or null`2662926630                        Two-letter ISO 3166-1 country code, uppercase.2663126632                      - `region: optional string or null`2663326634                        Region or state name.2663526636                        minLength: 1, maxLength: 2552663726638                      - `timezone: optional string or null`2663926640                        IANA timezone identifier, e.g. "America/Los_Angeles".2664126642                        minLength: 1, maxLength: 2552664326644                - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`2664526646                  Resolved default configuration for agent tools.2664726648                  - `enabled: boolean`2664926650                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2665126652                    Permission policy for tool execution.2665326654                    - `BetaManagedAgentsAlwaysAllowPolicy object`2665526656                      Tool calls are automatically approved without user confirmation.2665726658                    - `BetaManagedAgentsAlwaysAskPolicy object`2665926660                      Tool calls require user confirmation before execution.2666126662                    - `BetaManagedAgentsAutoPolicy object`2666326664                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2666526666              - `BetaManagedAgentsMCPToolset object`2666726668                - `type: "mcp_toolset"`2666926670                - `configs: array of BetaManagedAgentsMCPToolConfig`2667126672                  - `enabled: boolean`2667326674                  - `name: string`2667526676                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2667726678                    Permission policy for tool execution.2667926680                    - `BetaManagedAgentsAlwaysAllowPolicy object`2668126682                      Tool calls are automatically approved without user confirmation.2668326684                    - `BetaManagedAgentsAlwaysAskPolicy object`2668526686                      Tool calls require user confirmation before execution.2668726688                    - `BetaManagedAgentsAutoPolicy object`2668926690                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2669126692                - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`2669326694                  Resolved default configuration for all tools from an MCP server.2669526696                  - `enabled: boolean`2669726698                  - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`2669926700                    Permission policy for tool execution.2670126702                    - `BetaManagedAgentsAlwaysAllowPolicy object`2670326704                      Tool calls are automatically approved without user confirmation.2670526706                    - `BetaManagedAgentsAlwaysAskPolicy object`2670726708                      Tool calls require user confirmation before execution.2670926710                    - `BetaManagedAgentsAutoPolicy object`2671126712                      The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.2671326714                - `mcp_server_name: string`2671526716              - `BetaManagedAgentsCustomTool object`2671726718                A custom tool as returned in API responses.2671926720                - `type: "custom"`2672126722                - `description: string`2672326724                - `input_schema: BetaManagedAgentsCustomToolInputSchema`2672526726                  JSON Schema for custom tool input parameters.2672726728                  - `type: "object"`2672926730                  - `properties: optional map[unknown] or null`2673126732                  - `required: optional array of string or null`2673326734                - `name: string`2673526736            - `version: number`2673726738              format: int322673926740          - `BetaManagedAgentsAdvisor object`2674126742            Platform advisor roster entry: a model the session's primary thread may consult mid-turn.2674326744            - `type: "advisor"`2674526746            - `model: string`2674726748              The advisor model id.2674926750      - `name: string`2675126752      - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`2675326754        - `BetaManagedAgentsAnthropicSkill object`2675526756          A resolved Anthropic-managed skill.2675726758        - `BetaManagedAgentsCustomSkill object`2675926760          A resolved user-created custom skill.2676126762      - `system: string or null`2676326764      - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`2676526766        - `BetaManagedAgentsAgentToolset20260401 object`2676726768        - `BetaManagedAgentsMCPToolset object`2676926770        - `BetaManagedAgentsCustomTool object`2677126772          A custom tool as returned in API responses.2677326774      - `version: number`2677526776        format: int322677726778    - `budget: optional BetaManagedAgentsBudgetLimit or null`2677926780      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.2678126782      - `type: "limit"`2678326784      - `max_list_cost: BetaMonetaryAmount`2678526786        A monetary amount in a specific currency.2678726788        - `amount: string`2678926790          Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.2679126792        - `currency: BetaCurrency`2679326794          Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.2679526796    - `metadata: optional map[string]`2679726798      The session's full metadata bag after the update. Present when the update set non-empty metadata; absent when metadata was unchanged or cleared to empty.2679926800    - `title: optional string or null`2680126802      The session's new title. Present only when the update changed it.2680326804  - `BetaManagedAgentsStartEvent object`2680526806    Opens a preview of a buffered event. Carries the previewed event's type and id only. Followed by zero or more event_delta events with the same event id, normally concluded by the buffered event carrying that id. If the producing model request ends without that event (an error or interrupt mid-stream), its terminal span.model_request_end closes the preview. Only sent on stream connections that opt in via event_deltas; never appears in event history.2680726808    - `type: "event_start"`2680926810    - `event: BetaManagedAgentsStartEventPreview`2681126812      The previewed event's type and id. The event type determines which delta types the preview's event_delta events carry: agent.message events stream content_delta fragments; agent.thinking previews are start-only — no deltas follow, and the buffered agent.thinking with the same id concludes them.2681326814      - `BetaManagedAgentsAgentMessagePreview object`2681526816        - `type: "agent.message"`2681726818        - `id: string`2681926820          The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.2682126822      - `BetaManagedAgentsAgentThinkingPreview object`2682326824        - `type: "agent.thinking"`2682526826        - `id: string`2682726828          The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.2682926830  - `BetaManagedAgentsDeltaEvent object`2683126832    An incremental update to an event that is still being streamed. Deltas are best-effort and may stop early; when the buffered event with id == event_id is produced it carries the complete content. A model request that ends early (an error or interrupt) produces no buffered event — its terminal span.model_request_end closes the preview. Only sent on stream connections that opt in via event_deltas; never appears in event history.2683326834    - `type: "event_delta"`2683526836    - `delta: BetaManagedAgentsDeltaContent`2683726838      One fragment of the previewed event. The delta type is named for the previewed event's field it streams into: agent.message events stream content_delta fragments, each a partial element of the content array.2683926840      - `type: "content_delta"`2684126842      - `content: BetaManagedAgentsTextBlock`2684326844        Regular text content.2684526846      - `index: optional number`2684726848        Which entry in the previewed event's content array this fragment lands in. Insert content as that entry when the index is new; append to the existing entry otherwise.2684926850        format: uint322685126852    - `event_id: string`2685326854      The id of the event being previewed. Matches event.id on the corresponding event_start and the buffered event that reconciles the preview.2685526856  - `BetaManagedAgentsSystemMessageEvent object`2685726858    A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.2685926860    - `type: "system.message"`2686126862    - `id: string`2686326864      Unique identifier for this event.2686526866    - `content: array of BetaManagedAgentsSystemContentBlock`2686726868      System content blocks. Text-only.2686926870      - `type: "text"`2687126872      - `text: string`2687326874        The text content.2687526876        minLength: 12687726878    - `processed_at: optional string or null`2687926880      A timestamp in RFC 3339 format2688126882      format: date-time2688326884  - `BetaManagedAgentsSessionUsageEvent object`2688526886    Periodic snapshot of the session's cumulative usage and tracked list cost.2688726888    - `type: "session.usage"`2688926890    - `id: string`2689126892      Unique identifier for this event.2689326894    - `processed_at: string`2689526896      A timestamp in RFC 3339 format2689726898      format: date-time2689926900    - `usage: BetaManagedAgentsSessionUsageSnapshot`2690126902      Point-in-time snapshot of a session's cumulative usage.2690326904      - `active_seconds: optional number`2690526906        Cumulative time in seconds during which the session had at least one thread in running status. Overlapping activity from concurrent threads is counted once. This is the duration the session's runtime cost is priced on.2690726908        format: double2690926910      - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`2691126912        Prompt-cache creation token usage broken down by cache lifetime.2691326914        - `ephemeral_1h_input_tokens: optional number`2691526916          Tokens used to create 1-hour ephemeral cache entries.2691726918          format: int322691926920        - `ephemeral_5m_input_tokens: optional number`2692126922          Tokens used to create 5-minute ephemeral cache entries.2692326924          format: int322692526926      - `cache_read_input_tokens: optional number`2692726928        Total tokens read from prompt cache.2692926930        format: int322693126932      - `input_tokens: optional number`2693326934        Total input tokens consumed across all turns.2693526936        format: int322693726938      - `list_cost: optional BetaMonetaryAmount`2693926940        A monetary amount in a specific currency.2694126942      - `output_tokens: optional number`2694326944        Total output tokens generated across all turns.2694526946        format: int322694726948      - `server_tool_use: optional BetaManagedAgentsServerToolUsage`2694926950        Cumulative count of server-executed tool invocations, broken down by tool.2695126952        - `web_fetch_requests: optional number`2695326954          Number of server-executed web fetch requests.2695526956          format: int322695726958        - `web_search_requests: optional number`2695926960          Number of server-executed web search requests.2696126962          format: int322696326964    - `budget: optional BetaManagedAgentsBudgetLimit or null`2696526966      A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.2696726968#### Example2696926970```bash26971curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID/stream \26972    -H 'anthropic-version: 2023-06-01' \26973    -H 'anthropic-beta: managed-agents-2026-04-01' \26974    -H "X-Api-Key: $ANTHROPIC_API_KEY"26975```2697626977##### Response (200)2697826979```json26980{26981  "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",26982  "content": [26983    {26984      "text": "Where is my order #1234?",26985      "type": "text"26986    }26987  ],26988  "type": "user.message",26989  "processed_at": "2026-03-15T10:00:00Z"26990}26991```26992