Repository Public
1---2title: Threads3url: https://platform.claude.com/docs/en/api/beta/sessions/threads4---56# Threads78## List Session Threads910**GET** `/v1/sessions/{session_id}/threads`1112List Session Threads1314### Path parameters1516- `session_id: string`1718### Query parameters1920- `limit: optional number`2122 Maximum results per page. Defaults to 1000.2324 format: int322526- `page: optional string`2728 Opaque pagination cursor from a previous response's `next_page`. Forward-only.2930### Headers3132- `"anthropic-beta": optional array of AnthropicBeta`3334 Optional header to specify the beta version(s) you want to use.3536 - `string`3738 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`3940 - `"message-batches-2024-09-24"`4142 - `"prompt-caching-2024-07-31"`4344 - `"computer-use-2024-10-22"`4546 - `"computer-use-2025-01-24"`4748 - `"pdfs-2024-09-25"`4950 - `"token-counting-2024-11-01"`5152 - `"token-efficient-tools-2025-02-19"`5354 - `"output-128k-2025-02-19"`5556 - `"files-api-2025-04-14"`5758 - `"mcp-client-2025-04-04"`5960 - `"mcp-client-2025-11-20"`6162 - `"dev-full-thinking-2025-05-14"`6364 - `"interleaved-thinking-2025-05-14"`6566 - `"code-execution-2025-05-22"`6768 - `"extended-cache-ttl-2025-04-11"`6970 - `"context-1m-2025-08-07"`7172 - `"context-management-2025-06-27"`7374 - `"model-context-window-exceeded-2025-08-26"`7576 - `"skills-2025-10-02"`7778 - `"fast-mode-2026-02-01"`7980 - `"output-300k-2026-03-24"`8182 - `"user-profiles-2026-03-24"`8384 - `"user-profiles-2026-08-18"`8586 - `"user-profiles-2026-09-04"`8788 - `"advisor-tool-2026-03-01"`8990 - `"managed-agents-2026-04-01"`9192 - `"cache-diagnosis-2026-04-07"`9394 - `"dreaming-2026-04-21"`9596 - `"thinking-token-count-2026-05-13"`9798 - `"server-side-fallback-2026-06-01"`99100 - `"server-side-fallback-2026-07-01"`101102 - `"fallback-credit-2026-06-01"`103104 - `"fallback-credit-2026-07-01"`105106 - `"agent-memory-2026-07-22"`107108 - `"mid-conversation-tool-changes-2026-07-01"`109110 - `"compact-2026-01-12"`111112 - `"computer-use-2025-11-24"`113114 - `"mcp-tunnels-2026-06-22"`115116 - `"structured-outputs-2025-11-13"`117118 - `"task-budgets-2026-03-13"`119120 - `"thinking-display-updates-2026-08-18"`121122 - `"ce-user-management-2026-07-13"`123124 - `"mid-conversation-output-config-2026-07-01"`125126 - `"thinking-binding-controls-2026-08-01"`127128 - `"mid-conversation-system-clear-at-2026-08-21"`129130 - `"compact-2026-09-04"`131132- `"anthropic-workspace-id": optional string`133134### Returns135136- `data: optional array of BetaManagedAgentsSessionThread`137138 Threads in the session, primary first then children in spawn order.139140 - `type: "session_thread"`141142 - `id: string`143144 Unique identifier for this thread.145146 - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`147148 The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.149150 - `BetaManagedAgentsSessionThreadAgent object`151152 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`.153154 - `type: "agent"`155156 - `id: string`157158 - `description: string or null`159160 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`161162 - `type: "url"`163164 - `name: string`165166 - `url: string`167168 - `model: BetaManagedAgentsModelConfig`169170 Model identifier and configuration.171172 - `id: BetaManagedAgentsModel`173174 The model that will power your agent.175176 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.177178 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`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"`185186 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows187188 - `"claude-sonnet-5"`189190 High-performance model for coding and agents191192 - `"claude-fable-5"`193194 Next generation of intelligence for the hardest knowledge work and coding problems195196 - `"claude-opus-5"`197198 Powerful intelligence for long-running agents and coding199200 - `"claude-opus-4-8"`201202 Powerful intelligence for long-running agents and coding203204 - `"claude-opus-4-7"`205206 Powerful intelligence for long-running agents and coding207208 - `"claude-opus-4-6"`209210 Powerful intelligence for long-running agents and coding211212 - `"claude-sonnet-4-6"`213214 Best combination of speed and intelligence215216 - `"claude-haiku-4-5"`217218 Fastest model with near-frontier intelligence219220 - `"claude-haiku-4-5-20251001"`221222 Fastest model with near-frontier intelligence223224 - `"claude-opus-4-5"`225226 Powerful intelligence for long-running agents and coding227228 - `"claude-opus-4-5-20251101"`229230 Powerful intelligence for long-running agents and coding231232 - `"claude-sonnet-4-5"`233234 High-performance model for agents and coding235236 - `"claude-sonnet-4-5-20250929"`237238 High-performance model for agents and coding239240 - `string`241242 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`243244 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.245246 - `BetaManagedAgentsEffortLow object`247248 Low effort. Favors latency over reasoning depth.249250 - `type: "low"`251252 - `BetaManagedAgentsEffortMedium object`253254 Medium effort. Balances latency and reasoning depth.255256 - `type: "medium"`257258 - `BetaManagedAgentsEffortHigh object`259260 High effort. Favors reasoning depth.261262 - `type: "high"`263264 - `BetaManagedAgentsEffortXhigh object`265266 Extra-high effort. Not all models accept this level.267268 - `type: "xhigh"`269270 - `BetaManagedAgentsEffortMax object`271272 Maximum effort. Favors reasoning depth over latency.273274 - `type: "max"`275276 - `inference_geo: optional string`277278 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.279280 - `speed: optional "standard" or "fast"`281282 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.283284 - `"standard"`285286 - `"fast"`287288 - `name: string`289290 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`291292 - `BetaManagedAgentsAnthropicSkill object`293294 A resolved Anthropic-managed skill.295296 - `type: "anthropic"`297298 - `skill_id: string`299300 - `version: string`301302 - `BetaManagedAgentsCustomSkill object`303304 A resolved user-created custom skill.305306 - `type: "custom"`307308 - `skill_id: string`309310 - `version: string`311312 - `system: string or null`313314 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`315316 - `BetaManagedAgentsAgentToolset20260401 object`317318 - `type: "agent_toolset_20260401"`319320 - `configs: array of BetaManagedAgentsAgentToolConfig`321322 - `BetaManagedAgentsBashToolConfig object`323324 Configuration for the bash tool.325326 - `type: "bash"`327328 - `enabled: boolean`329330 - `name: "bash"`331332 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`333334 Permission policy for tool execution.335336 - `BetaManagedAgentsAlwaysAllowPolicy object`337338 Tool calls are automatically approved without user confirmation.339340 - `type: "always_allow"`341342 - `BetaManagedAgentsAlwaysAskPolicy object`343344 Tool calls require user confirmation before execution.345346 - `type: "always_ask"`347348 - `BetaManagedAgentsAutoPolicy object`349350 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.351352 - `type: "auto"`353354 - `BetaManagedAgentsEditToolConfig object`355356 Configuration for the edit tool.357358 - `type: "edit"`359360 - `enabled: boolean`361362 - `name: "edit"`363364 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`365366 Permission policy for tool execution.367368 - `BetaManagedAgentsAlwaysAllowPolicy object`369370 Tool calls are automatically approved without user confirmation.371372 - `BetaManagedAgentsAlwaysAskPolicy object`373374 Tool calls require user confirmation before execution.375376 - `BetaManagedAgentsAutoPolicy object`377378 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.379380 - `BetaManagedAgentsReadToolConfig object`381382 Configuration for the read tool.383384 - `type: "read"`385386 - `enabled: boolean`387388 - `name: "read"`389390 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`391392 Permission policy for tool execution.393394 - `BetaManagedAgentsAlwaysAllowPolicy object`395396 Tool calls are automatically approved without user confirmation.397398 - `BetaManagedAgentsAlwaysAskPolicy object`399400 Tool calls require user confirmation before execution.401402 - `BetaManagedAgentsAutoPolicy object`403404 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.405406 - `BetaManagedAgentsWriteToolConfig object`407408 Configuration for the write tool.409410 - `type: "write"`411412 - `enabled: boolean`413414 - `name: "write"`415416 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`417418 Permission policy for tool execution.419420 - `BetaManagedAgentsAlwaysAllowPolicy object`421422 Tool calls are automatically approved without user confirmation.423424 - `BetaManagedAgentsAlwaysAskPolicy object`425426 Tool calls require user confirmation before execution.427428 - `BetaManagedAgentsAutoPolicy object`429430 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.431432 - `BetaManagedAgentsGlobToolConfig object`433434 Configuration for the glob tool.435436 - `type: "glob"`437438 - `enabled: boolean`439440 - `name: "glob"`441442 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`443444 Permission policy for tool execution.445446 - `BetaManagedAgentsAlwaysAllowPolicy object`447448 Tool calls are automatically approved without user confirmation.449450 - `BetaManagedAgentsAlwaysAskPolicy object`451452 Tool calls require user confirmation before execution.453454 - `BetaManagedAgentsAutoPolicy object`455456 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.457458 - `BetaManagedAgentsGrepToolConfig object`459460 Configuration for the grep tool.461462 - `type: "grep"`463464 - `enabled: boolean`465466 - `name: "grep"`467468 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`469470 Permission policy for tool execution.471472 - `BetaManagedAgentsAlwaysAllowPolicy object`473474 Tool calls are automatically approved without user confirmation.475476 - `BetaManagedAgentsAlwaysAskPolicy object`477478 Tool calls require user confirmation before execution.479480 - `BetaManagedAgentsAutoPolicy object`481482 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.483484 - `BetaManagedAgentsWebFetchToolConfig object`485486 Configuration for the web_fetch tool.487488 - `type: "web_fetch"`489490 - `enabled: boolean`491492 - `name: "web_fetch"`493494 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`495496 Permission policy for tool execution.497498 - `BetaManagedAgentsAlwaysAllowPolicy object`499500 Tool calls are automatically approved without user confirmation.501502 - `BetaManagedAgentsAlwaysAskPolicy object`503504 Tool calls require user confirmation before execution.505506 - `BetaManagedAgentsAutoPolicy object`507508 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.509510 - `allowed_domains: optional array of string`511512 - `blocked_domains: optional array of string`513514 - `max_content_tokens: optional number or null`515516 format: int32517518 - `BetaManagedAgentsWebSearchToolConfig object`519520 Configuration for the web_search tool.521522 - `type: "web_search"`523524 - `enabled: boolean`525526 - `name: "web_search"`527528 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`529530 Permission policy for tool execution.531532 - `BetaManagedAgentsAlwaysAllowPolicy object`533534 Tool calls are automatically approved without user confirmation.535536 - `BetaManagedAgentsAlwaysAskPolicy object`537538 Tool calls require user confirmation before execution.539540 - `BetaManagedAgentsAutoPolicy object`541542 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.543544 - `allowed_domains: optional array of string`545546 - `blocked_domains: optional array of string`547548 - `user_location: optional BetaManagedAgentsUserLocation or null`549550 Approximate user location for search result localization.551552 - `type: "approximate"`553554 Location precision. Only "approximate" is supported.555556 - `city: optional string or null`557558 City name.559560 minLength: 1, maxLength: 255561562 - `country: optional string or null`563564 Two-letter ISO 3166-1 country code, uppercase.565566 - `region: optional string or null`567568 Region or state name.569570 minLength: 1, maxLength: 255571572 - `timezone: optional string or null`573574 IANA timezone identifier, e.g. "America/Los_Angeles".575576 minLength: 1, maxLength: 255577578 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`579580 Resolved default configuration for agent tools.581582 - `enabled: boolean`583584 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`585586 Permission policy for tool execution.587588 - `BetaManagedAgentsAlwaysAllowPolicy object`589590 Tool calls are automatically approved without user confirmation.591592 - `BetaManagedAgentsAlwaysAskPolicy object`593594 Tool calls require user confirmation before execution.595596 - `BetaManagedAgentsAutoPolicy object`597598 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.599600 - `BetaManagedAgentsMCPToolset object`601602 - `type: "mcp_toolset"`603604 - `configs: array of BetaManagedAgentsMCPToolConfig`605606 - `enabled: boolean`607608 - `name: string`609610 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`611612 Permission policy for tool execution.613614 - `BetaManagedAgentsAlwaysAllowPolicy object`615616 Tool calls are automatically approved without user confirmation.617618 - `BetaManagedAgentsAlwaysAskPolicy object`619620 Tool calls require user confirmation before execution.621622 - `BetaManagedAgentsAutoPolicy object`623624 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.625626 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`627628 Resolved default configuration for all tools from an MCP server.629630 - `enabled: boolean`631632 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`633634 Permission policy for tool execution.635636 - `BetaManagedAgentsAlwaysAllowPolicy object`637638 Tool calls are automatically approved without user confirmation.639640 - `BetaManagedAgentsAlwaysAskPolicy object`641642 Tool calls require user confirmation before execution.643644 - `BetaManagedAgentsAutoPolicy object`645646 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.647648 - `mcp_server_name: string`649650 - `BetaManagedAgentsCustomTool object`651652 A custom tool as returned in API responses.653654 - `type: "custom"`655656 - `description: string`657658 - `input_schema: BetaManagedAgentsCustomToolInputSchema`659660 JSON Schema for custom tool input parameters.661662 - `type: "object"`663664 - `properties: optional map[unknown] or null`665666 - `required: optional array of string or null`667668 - `name: string`669670 - `version: number`671672 format: int32673674 - `BetaManagedAgentsAdvisor object`675676 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.677678 - `type: "advisor"`679680 - `model: string`681682 The advisor model id.683684 - `archived_at: string or null`685686 A timestamp in RFC 3339 format687688 format: date-time689690 - `created_at: string`691692 A timestamp in RFC 3339 format693694 format: date-time695696 - `parent_thread_id: string or null`697698 Parent thread that spawned this thread. Null for the primary thread.699700 - `session_id: string`701702 The session this thread belongs to.703704 - `stats: BetaManagedAgentsSessionThreadStats or null`705706 Timing statistics for a session thread.707708 - `active_seconds: optional number`709710 Cumulative time in seconds the thread spent actively running. Excludes idle time.711712 format: double713714 - `duration_seconds: optional number`715716 Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.717718 format: double719720 - `startup_seconds: optional number`721722 Time in seconds for the thread to begin running. Zero for child threads, which start immediately.723724 format: double725726 - `status: BetaManagedAgentsSessionThreadStatus`727728 SessionThreadStatus enum729730 - `"running"`731732 - `"idle"`733734 - `"rescheduling"`735736 - `"terminated"`737738 - `updated_at: string`739740 A timestamp in RFC 3339 format741742 format: date-time743744 - `usage: BetaManagedAgentsSessionThreadUsage or null`745746 Cumulative token usage for a session thread across all turns.747748 - `active_seconds: optional number`749750 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.751752 format: double753754 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`755756 Prompt-cache creation token usage broken down by cache lifetime.757758 - `ephemeral_1h_input_tokens: optional number`759760 Tokens used to create 1-hour ephemeral cache entries.761762 format: int32763764 - `ephemeral_5m_input_tokens: optional number`765766 Tokens used to create 5-minute ephemeral cache entries.767768 format: int32769770 - `cache_read_input_tokens: optional number`771772 Total tokens read from prompt cache.773774 format: int32775776 - `input_tokens: optional number`777778 Total input tokens consumed across all turns.779780 format: int32781782 - `list_cost: optional BetaMonetaryAmount or null`783784 A monetary amount in a specific currency.785786 - `amount: string`787788 Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.789790 - `currency: BetaCurrency`791792 Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.793794 - `output_tokens: optional number`795796 Total output tokens generated across all turns.797798 format: int32799800 - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`801802 Cumulative count of server-executed tool invocations, broken down by tool.803804 - `web_fetch_requests: optional number`805806 Number of server-executed web fetch requests.807808 format: int32809810 - `web_search_requests: optional number`811812 Number of server-executed web search requests.813814 format: int32815816- `next_page: optional string or null`817818 Opaque cursor for the next page. Null when no more results.819820### Example821822```bash823curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads \824 -H 'anthropic-version: 2023-06-01' \825 -H 'anthropic-beta: managed-agents-2026-04-01' \826 -H "X-Api-Key: $ANTHROPIC_API_KEY"827```828829#### Response (200)830831```json832{833 "data": [834 {835 "id": "sthr_011CZkZVWa6oIjw0rgXZpnBt",836 "agent": {837 "id": "agent_011CZkYqphY8vELVzwCUpqiQ",838 "description": "A focused research subagent.",839 "mcp_servers": [840 {841 "name": "example-mcp",842 "type": "url",843 "url": "https://example-server.modelcontextprotocol.io/sse"844 }845 ],846 "model": {847 "id": "claude-opus-5",848 "effort": {849 "type": "low"850 },851 "inference_geo": "inference_geo",852 "speed": "standard"853 },854 "name": "Researcher",855 "skills": [856 {857 "skill_id": "xlsx",858 "type": "anthropic",859 "version": "1"860 }861 ],862 "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",863 "tools": [864 {865 "configs": [866 {867 "enabled": true,868 "name": "bash",869 "permission_policy": {870 "type": "always_allow"871 },872 "type": "bash"873 }874 ],875 "default_config": {876 "enabled": true,877 "permission_policy": {878 "type": "always_ask"879 }880 },881 "type": "agent_toolset_20260401"882 }883 ],884 "type": "agent",885 "version": 1886 },887 "archived_at": null,888 "created_at": "2026-03-15T10:00:00Z",889 "parent_thread_id": null,890 "session_id": "sesn_011CZkZAtmR3yMPDzynEDxu7",891 "stats": {892 "active_seconds": 0,893 "duration_seconds": 0,894 "startup_seconds": 0895 },896 "status": "idle",897 "type": "session_thread",898 "updated_at": "2026-03-15T10:00:00Z",899 "usage": {900 "active_seconds": 0,901 "cache_creation": {902 "ephemeral_1h_input_tokens": 0,903 "ephemeral_5m_input_tokens": 0904 },905 "cache_read_input_tokens": 0,906 "input_tokens": 0,907 "list_cost": {908 "amount": "2500",909 "currency": "USD"910 },911 "output_tokens": 0,912 "server_tool_use": {913 "web_fetch_requests": 0,914 "web_search_requests": 3915 }916 }917 }918 ],919 "next_page": "page_MjAyNS0wNS0xNFQwMDowMDowMFo="920}921```922923## Get Session Thread924925**GET** `/v1/sessions/{session_id}/threads/{thread_id}`926927Get Session Thread928929### Path parameters930931- `session_id: string`932933- `thread_id: string`934935### Headers936937- `"anthropic-beta": optional array of AnthropicBeta`938939 Optional header to specify the beta version(s) you want to use.940941 - `string`942943 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`944945 - `"message-batches-2024-09-24"`946947 - `"prompt-caching-2024-07-31"`948949 - `"computer-use-2024-10-22"`950951 - `"computer-use-2025-01-24"`952953 - `"pdfs-2024-09-25"`954955 - `"token-counting-2024-11-01"`956957 - `"token-efficient-tools-2025-02-19"`958959 - `"output-128k-2025-02-19"`960961 - `"files-api-2025-04-14"`962963 - `"mcp-client-2025-04-04"`964965 - `"mcp-client-2025-11-20"`966967 - `"dev-full-thinking-2025-05-14"`968969 - `"interleaved-thinking-2025-05-14"`970971 - `"code-execution-2025-05-22"`972973 - `"extended-cache-ttl-2025-04-11"`974975 - `"context-1m-2025-08-07"`976977 - `"context-management-2025-06-27"`978979 - `"model-context-window-exceeded-2025-08-26"`980981 - `"skills-2025-10-02"`982983 - `"fast-mode-2026-02-01"`984985 - `"output-300k-2026-03-24"`986987 - `"user-profiles-2026-03-24"`988989 - `"user-profiles-2026-08-18"`990991 - `"user-profiles-2026-09-04"`992993 - `"advisor-tool-2026-03-01"`994995 - `"managed-agents-2026-04-01"`996997 - `"cache-diagnosis-2026-04-07"`998999 - `"dreaming-2026-04-21"`10001001 - `"thinking-token-count-2026-05-13"`10021003 - `"server-side-fallback-2026-06-01"`10041005 - `"server-side-fallback-2026-07-01"`10061007 - `"fallback-credit-2026-06-01"`10081009 - `"fallback-credit-2026-07-01"`10101011 - `"agent-memory-2026-07-22"`10121013 - `"mid-conversation-tool-changes-2026-07-01"`10141015 - `"compact-2026-01-12"`10161017 - `"computer-use-2025-11-24"`10181019 - `"mcp-tunnels-2026-06-22"`10201021 - `"structured-outputs-2025-11-13"`10221023 - `"task-budgets-2026-03-13"`10241025 - `"thinking-display-updates-2026-08-18"`10261027 - `"ce-user-management-2026-07-13"`10281029 - `"mid-conversation-output-config-2026-07-01"`10301031 - `"thinking-binding-controls-2026-08-01"`10321033 - `"mid-conversation-system-clear-at-2026-08-21"`10341035 - `"compact-2026-09-04"`10361037- `"anthropic-workspace-id": optional string`10381039### Returns10401041- `BetaManagedAgentsSessionThread object`10421043 An execution thread within a `session`. Each session has one primary thread plus zero or more child threads spawned by the coordinator.10441045 - `type: "session_thread"`10461047 - `id: string`10481049 Unique identifier for this thread.10501051 - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`10521053 The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.10541055 - `BetaManagedAgentsSessionThreadAgent object`10561057 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`.10581059 - `type: "agent"`10601061 - `id: string`10621063 - `description: string or null`10641065 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`10661067 - `type: "url"`10681069 - `name: string`10701071 - `url: string`10721073 - `model: BetaManagedAgentsModelConfig`10741075 Model identifier and configuration.10761077 - `id: BetaManagedAgentsModel`10781079 The model that will power your agent.10801081 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.10821083 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`10841085 The model that will power your agent.10861087 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.10881089 - `"claude-fable-5-1"`10901091 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows10921093 - `"claude-sonnet-5"`10941095 High-performance model for coding and agents10961097 - `"claude-fable-5"`10981099 Next generation of intelligence for the hardest knowledge work and coding problems11001101 - `"claude-opus-5"`11021103 Powerful intelligence for long-running agents and coding11041105 - `"claude-opus-4-8"`11061107 Powerful intelligence for long-running agents and coding11081109 - `"claude-opus-4-7"`11101111 Powerful intelligence for long-running agents and coding11121113 - `"claude-opus-4-6"`11141115 Powerful intelligence for long-running agents and coding11161117 - `"claude-sonnet-4-6"`11181119 Best combination of speed and intelligence11201121 - `"claude-haiku-4-5"`11221123 Fastest model with near-frontier intelligence11241125 - `"claude-haiku-4-5-20251001"`11261127 Fastest model with near-frontier intelligence11281129 - `"claude-opus-4-5"`11301131 Powerful intelligence for long-running agents and coding11321133 - `"claude-opus-4-5-20251101"`11341135 Powerful intelligence for long-running agents and coding11361137 - `"claude-sonnet-4-5"`11381139 High-performance model for agents and coding11401141 - `"claude-sonnet-4-5-20250929"`11421143 High-performance model for agents and coding11441145 - `string`11461147 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`11481149 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.11501151 - `BetaManagedAgentsEffortLow object`11521153 Low effort. Favors latency over reasoning depth.11541155 - `type: "low"`11561157 - `BetaManagedAgentsEffortMedium object`11581159 Medium effort. Balances latency and reasoning depth.11601161 - `type: "medium"`11621163 - `BetaManagedAgentsEffortHigh object`11641165 High effort. Favors reasoning depth.11661167 - `type: "high"`11681169 - `BetaManagedAgentsEffortXhigh object`11701171 Extra-high effort. Not all models accept this level.11721173 - `type: "xhigh"`11741175 - `BetaManagedAgentsEffortMax object`11761177 Maximum effort. Favors reasoning depth over latency.11781179 - `type: "max"`11801181 - `inference_geo: optional string`11821183 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.11841185 - `speed: optional "standard" or "fast"`11861187 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.11881189 - `"standard"`11901191 - `"fast"`11921193 - `name: string`11941195 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`11961197 - `BetaManagedAgentsAnthropicSkill object`11981199 A resolved Anthropic-managed skill.12001201 - `type: "anthropic"`12021203 - `skill_id: string`12041205 - `version: string`12061207 - `BetaManagedAgentsCustomSkill object`12081209 A resolved user-created custom skill.12101211 - `type: "custom"`12121213 - `skill_id: string`12141215 - `version: string`12161217 - `system: string or null`12181219 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`12201221 - `BetaManagedAgentsAgentToolset20260401 object`12221223 - `type: "agent_toolset_20260401"`12241225 - `configs: array of BetaManagedAgentsAgentToolConfig`12261227 - `BetaManagedAgentsBashToolConfig object`12281229 Configuration for the bash tool.12301231 - `type: "bash"`12321233 - `enabled: boolean`12341235 - `name: "bash"`12361237 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`12381239 Permission policy for tool execution.12401241 - `BetaManagedAgentsAlwaysAllowPolicy object`12421243 Tool calls are automatically approved without user confirmation.12441245 - `type: "always_allow"`12461247 - `BetaManagedAgentsAlwaysAskPolicy object`12481249 Tool calls require user confirmation before execution.12501251 - `type: "always_ask"`12521253 - `BetaManagedAgentsAutoPolicy object`12541255 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.12561257 - `type: "auto"`12581259 - `BetaManagedAgentsEditToolConfig object`12601261 Configuration for the edit tool.12621263 - `type: "edit"`12641265 - `enabled: boolean`12661267 - `name: "edit"`12681269 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`12701271 Permission policy for tool execution.12721273 - `BetaManagedAgentsAlwaysAllowPolicy object`12741275 Tool calls are automatically approved without user confirmation.12761277 - `BetaManagedAgentsAlwaysAskPolicy object`12781279 Tool calls require user confirmation before execution.12801281 - `BetaManagedAgentsAutoPolicy object`12821283 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.12841285 - `BetaManagedAgentsReadToolConfig object`12861287 Configuration for the read tool.12881289 - `type: "read"`12901291 - `enabled: boolean`12921293 - `name: "read"`12941295 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`12961297 Permission policy for tool execution.12981299 - `BetaManagedAgentsAlwaysAllowPolicy object`13001301 Tool calls are automatically approved without user confirmation.13021303 - `BetaManagedAgentsAlwaysAskPolicy object`13041305 Tool calls require user confirmation before execution.13061307 - `BetaManagedAgentsAutoPolicy object`13081309 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.13101311 - `BetaManagedAgentsWriteToolConfig object`13121313 Configuration for the write tool.13141315 - `type: "write"`13161317 - `enabled: boolean`13181319 - `name: "write"`13201321 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`13221323 Permission policy for tool execution.13241325 - `BetaManagedAgentsAlwaysAllowPolicy object`13261327 Tool calls are automatically approved without user confirmation.13281329 - `BetaManagedAgentsAlwaysAskPolicy object`13301331 Tool calls require user confirmation before execution.13321333 - `BetaManagedAgentsAutoPolicy object`13341335 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.13361337 - `BetaManagedAgentsGlobToolConfig object`13381339 Configuration for the glob tool.13401341 - `type: "glob"`13421343 - `enabled: boolean`13441345 - `name: "glob"`13461347 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`13481349 Permission policy for tool execution.13501351 - `BetaManagedAgentsAlwaysAllowPolicy object`13521353 Tool calls are automatically approved without user confirmation.13541355 - `BetaManagedAgentsAlwaysAskPolicy object`13561357 Tool calls require user confirmation before execution.13581359 - `BetaManagedAgentsAutoPolicy object`13601361 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.13621363 - `BetaManagedAgentsGrepToolConfig object`13641365 Configuration for the grep tool.13661367 - `type: "grep"`13681369 - `enabled: boolean`13701371 - `name: "grep"`13721373 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`13741375 Permission policy for tool execution.13761377 - `BetaManagedAgentsAlwaysAllowPolicy object`13781379 Tool calls are automatically approved without user confirmation.13801381 - `BetaManagedAgentsAlwaysAskPolicy object`13821383 Tool calls require user confirmation before execution.13841385 - `BetaManagedAgentsAutoPolicy object`13861387 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.13881389 - `BetaManagedAgentsWebFetchToolConfig object`13901391 Configuration for the web_fetch tool.13921393 - `type: "web_fetch"`13941395 - `enabled: boolean`13961397 - `name: "web_fetch"`13981399 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14001401 Permission policy for tool execution.14021403 - `BetaManagedAgentsAlwaysAllowPolicy object`14041405 Tool calls are automatically approved without user confirmation.14061407 - `BetaManagedAgentsAlwaysAskPolicy object`14081409 Tool calls require user confirmation before execution.14101411 - `BetaManagedAgentsAutoPolicy object`14121413 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.14141415 - `allowed_domains: optional array of string`14161417 - `blocked_domains: optional array of string`14181419 - `max_content_tokens: optional number or null`14201421 format: int3214221423 - `BetaManagedAgentsWebSearchToolConfig object`14241425 Configuration for the web_search tool.14261427 - `type: "web_search"`14281429 - `enabled: boolean`14301431 - `name: "web_search"`14321433 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14341435 Permission policy for tool execution.14361437 - `BetaManagedAgentsAlwaysAllowPolicy object`14381439 Tool calls are automatically approved without user confirmation.14401441 - `BetaManagedAgentsAlwaysAskPolicy object`14421443 Tool calls require user confirmation before execution.14441445 - `BetaManagedAgentsAutoPolicy object`14461447 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.14481449 - `allowed_domains: optional array of string`14501451 - `blocked_domains: optional array of string`14521453 - `user_location: optional BetaManagedAgentsUserLocation or null`14541455 Approximate user location for search result localization.14561457 - `type: "approximate"`14581459 Location precision. Only "approximate" is supported.14601461 - `city: optional string or null`14621463 City name.14641465 minLength: 1, maxLength: 25514661467 - `country: optional string or null`14681469 Two-letter ISO 3166-1 country code, uppercase.14701471 - `region: optional string or null`14721473 Region or state name.14741475 minLength: 1, maxLength: 25514761477 - `timezone: optional string or null`14781479 IANA timezone identifier, e.g. "America/Los_Angeles".14801481 minLength: 1, maxLength: 25514821483 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`14841485 Resolved default configuration for agent tools.14861487 - `enabled: boolean`14881489 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`14901491 Permission policy for tool execution.14921493 - `BetaManagedAgentsAlwaysAllowPolicy object`14941495 Tool calls are automatically approved without user confirmation.14961497 - `BetaManagedAgentsAlwaysAskPolicy object`14981499 Tool calls require user confirmation before execution.15001501 - `BetaManagedAgentsAutoPolicy object`15021503 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.15041505 - `BetaManagedAgentsMCPToolset object`15061507 - `type: "mcp_toolset"`15081509 - `configs: array of BetaManagedAgentsMCPToolConfig`15101511 - `enabled: boolean`15121513 - `name: string`15141515 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`15161517 Permission policy for tool execution.15181519 - `BetaManagedAgentsAlwaysAllowPolicy object`15201521 Tool calls are automatically approved without user confirmation.15221523 - `BetaManagedAgentsAlwaysAskPolicy object`15241525 Tool calls require user confirmation before execution.15261527 - `BetaManagedAgentsAutoPolicy object`15281529 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.15301531 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`15321533 Resolved default configuration for all tools from an MCP server.15341535 - `enabled: boolean`15361537 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`15381539 Permission policy for tool execution.15401541 - `BetaManagedAgentsAlwaysAllowPolicy object`15421543 Tool calls are automatically approved without user confirmation.15441545 - `BetaManagedAgentsAlwaysAskPolicy object`15461547 Tool calls require user confirmation before execution.15481549 - `BetaManagedAgentsAutoPolicy object`15501551 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.15521553 - `mcp_server_name: string`15541555 - `BetaManagedAgentsCustomTool object`15561557 A custom tool as returned in API responses.15581559 - `type: "custom"`15601561 - `description: string`15621563 - `input_schema: BetaManagedAgentsCustomToolInputSchema`15641565 JSON Schema for custom tool input parameters.15661567 - `type: "object"`15681569 - `properties: optional map[unknown] or null`15701571 - `required: optional array of string or null`15721573 - `name: string`15741575 - `version: number`15761577 format: int3215781579 - `BetaManagedAgentsAdvisor object`15801581 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.15821583 - `type: "advisor"`15841585 - `model: string`15861587 The advisor model id.15881589 - `archived_at: string or null`15901591 A timestamp in RFC 3339 format15921593 format: date-time15941595 - `created_at: string`15961597 A timestamp in RFC 3339 format15981599 format: date-time16001601 - `parent_thread_id: string or null`16021603 Parent thread that spawned this thread. Null for the primary thread.16041605 - `session_id: string`16061607 The session this thread belongs to.16081609 - `stats: BetaManagedAgentsSessionThreadStats or null`16101611 Timing statistics for a session thread.16121613 - `active_seconds: optional number`16141615 Cumulative time in seconds the thread spent actively running. Excludes idle time.16161617 format: double16181619 - `duration_seconds: optional number`16201621 Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.16221623 format: double16241625 - `startup_seconds: optional number`16261627 Time in seconds for the thread to begin running. Zero for child threads, which start immediately.16281629 format: double16301631 - `status: BetaManagedAgentsSessionThreadStatus`16321633 SessionThreadStatus enum16341635 - `"running"`16361637 - `"idle"`16381639 - `"rescheduling"`16401641 - `"terminated"`16421643 - `updated_at: string`16441645 A timestamp in RFC 3339 format16461647 format: date-time16481649 - `usage: BetaManagedAgentsSessionThreadUsage or null`16501651 Cumulative token usage for a session thread across all turns.16521653 - `active_seconds: optional number`16541655 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.16561657 format: double16581659 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`16601661 Prompt-cache creation token usage broken down by cache lifetime.16621663 - `ephemeral_1h_input_tokens: optional number`16641665 Tokens used to create 1-hour ephemeral cache entries.16661667 format: int3216681669 - `ephemeral_5m_input_tokens: optional number`16701671 Tokens used to create 5-minute ephemeral cache entries.16721673 format: int3216741675 - `cache_read_input_tokens: optional number`16761677 Total tokens read from prompt cache.16781679 format: int3216801681 - `input_tokens: optional number`16821683 Total input tokens consumed across all turns.16841685 format: int3216861687 - `list_cost: optional BetaMonetaryAmount or null`16881689 A monetary amount in a specific currency.16901691 - `amount: string`16921693 Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.16941695 - `currency: BetaCurrency`16961697 Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.16981699 - `output_tokens: optional number`17001701 Total output tokens generated across all turns.17021703 format: int3217041705 - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`17061707 Cumulative count of server-executed tool invocations, broken down by tool.17081709 - `web_fetch_requests: optional number`17101711 Number of server-executed web fetch requests.17121713 format: int3217141715 - `web_search_requests: optional number`17161717 Number of server-executed web search requests.17181719 format: int3217201721### Example17221723```bash1724curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID \1725 -H 'anthropic-version: 2023-06-01' \1726 -H 'anthropic-beta: managed-agents-2026-04-01' \1727 -H "X-Api-Key: $ANTHROPIC_API_KEY"1728```17291730#### Response (200)17311732```json1733{1734 "id": "sthr_011CZkZVWa6oIjw0rgXZpnBt",1735 "agent": {1736 "id": "agent_011CZkYqphY8vELVzwCUpqiQ",1737 "description": "A focused research subagent.",1738 "mcp_servers": [1739 {1740 "name": "example-mcp",1741 "type": "url",1742 "url": "https://example-server.modelcontextprotocol.io/sse"1743 }1744 ],1745 "model": {1746 "id": "claude-opus-5",1747 "effort": {1748 "type": "low"1749 },1750 "inference_geo": "inference_geo",1751 "speed": "standard"1752 },1753 "name": "Researcher",1754 "skills": [1755 {1756 "skill_id": "xlsx",1757 "type": "anthropic",1758 "version": "1"1759 }1760 ],1761 "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",1762 "tools": [1763 {1764 "configs": [1765 {1766 "enabled": true,1767 "name": "bash",1768 "permission_policy": {1769 "type": "always_allow"1770 },1771 "type": "bash"1772 }1773 ],1774 "default_config": {1775 "enabled": true,1776 "permission_policy": {1777 "type": "always_ask"1778 }1779 },1780 "type": "agent_toolset_20260401"1781 }1782 ],1783 "type": "agent",1784 "version": 11785 },1786 "archived_at": null,1787 "created_at": "2026-03-15T10:00:00Z",1788 "parent_thread_id": null,1789 "session_id": "sesn_011CZkZAtmR3yMPDzynEDxu7",1790 "stats": {1791 "active_seconds": 0,1792 "duration_seconds": 0,1793 "startup_seconds": 01794 },1795 "status": "idle",1796 "type": "session_thread",1797 "updated_at": "2026-03-15T10:00:00Z",1798 "usage": {1799 "active_seconds": 0,1800 "cache_creation": {1801 "ephemeral_1h_input_tokens": 0,1802 "ephemeral_5m_input_tokens": 01803 },1804 "cache_read_input_tokens": 0,1805 "input_tokens": 0,1806 "list_cost": {1807 "amount": "2500",1808 "currency": "USD"1809 },1810 "output_tokens": 0,1811 "server_tool_use": {1812 "web_fetch_requests": 0,1813 "web_search_requests": 31814 }1815 }1816}1817```18181819## Archive Session Thread18201821**POST** `/v1/sessions/{session_id}/threads/{thread_id}/archive`18221823Archive Session Thread18241825### Path parameters18261827- `session_id: string`18281829- `thread_id: string`18301831### Headers18321833- `"anthropic-beta": optional array of AnthropicBeta`18341835 Optional header to specify the beta version(s) you want to use.18361837 - `string`18381839 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`18401841 - `"message-batches-2024-09-24"`18421843 - `"prompt-caching-2024-07-31"`18441845 - `"computer-use-2024-10-22"`18461847 - `"computer-use-2025-01-24"`18481849 - `"pdfs-2024-09-25"`18501851 - `"token-counting-2024-11-01"`18521853 - `"token-efficient-tools-2025-02-19"`18541855 - `"output-128k-2025-02-19"`18561857 - `"files-api-2025-04-14"`18581859 - `"mcp-client-2025-04-04"`18601861 - `"mcp-client-2025-11-20"`18621863 - `"dev-full-thinking-2025-05-14"`18641865 - `"interleaved-thinking-2025-05-14"`18661867 - `"code-execution-2025-05-22"`18681869 - `"extended-cache-ttl-2025-04-11"`18701871 - `"context-1m-2025-08-07"`18721873 - `"context-management-2025-06-27"`18741875 - `"model-context-window-exceeded-2025-08-26"`18761877 - `"skills-2025-10-02"`18781879 - `"fast-mode-2026-02-01"`18801881 - `"output-300k-2026-03-24"`18821883 - `"user-profiles-2026-03-24"`18841885 - `"user-profiles-2026-08-18"`18861887 - `"user-profiles-2026-09-04"`18881889 - `"advisor-tool-2026-03-01"`18901891 - `"managed-agents-2026-04-01"`18921893 - `"cache-diagnosis-2026-04-07"`18941895 - `"dreaming-2026-04-21"`18961897 - `"thinking-token-count-2026-05-13"`18981899 - `"server-side-fallback-2026-06-01"`19001901 - `"server-side-fallback-2026-07-01"`19021903 - `"fallback-credit-2026-06-01"`19041905 - `"fallback-credit-2026-07-01"`19061907 - `"agent-memory-2026-07-22"`19081909 - `"mid-conversation-tool-changes-2026-07-01"`19101911 - `"compact-2026-01-12"`19121913 - `"computer-use-2025-11-24"`19141915 - `"mcp-tunnels-2026-06-22"`19161917 - `"structured-outputs-2025-11-13"`19181919 - `"task-budgets-2026-03-13"`19201921 - `"thinking-display-updates-2026-08-18"`19221923 - `"ce-user-management-2026-07-13"`19241925 - `"mid-conversation-output-config-2026-07-01"`19261927 - `"thinking-binding-controls-2026-08-01"`19281929 - `"mid-conversation-system-clear-at-2026-08-21"`19301931 - `"compact-2026-09-04"`19321933- `"anthropic-workspace-id": optional string`19341935### Returns19361937- `BetaManagedAgentsSessionThread object`19381939 An execution thread within a `session`. Each session has one primary thread plus zero or more child threads spawned by the coordinator.19401941 - `type: "session_thread"`19421943 - `id: string`19441945 Unique identifier for this thread.19461947 - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`19481949 The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.19501951 - `BetaManagedAgentsSessionThreadAgent object`19521953 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`.19541955 - `type: "agent"`19561957 - `id: string`19581959 - `description: string or null`19601961 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`19621963 - `type: "url"`19641965 - `name: string`19661967 - `url: string`19681969 - `model: BetaManagedAgentsModelConfig`19701971 Model identifier and configuration.19721973 - `id: BetaManagedAgentsModel`19741975 The model that will power your agent.19761977 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.19781979 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`19801981 The model that will power your agent.19821983 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.19841985 - `"claude-fable-5-1"`19861987 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows19881989 - `"claude-sonnet-5"`19901991 High-performance model for coding and agents19921993 - `"claude-fable-5"`19941995 Next generation of intelligence for the hardest knowledge work and coding problems19961997 - `"claude-opus-5"`19981999 Powerful intelligence for long-running agents and coding20002001 - `"claude-opus-4-8"`20022003 Powerful intelligence for long-running agents and coding20042005 - `"claude-opus-4-7"`20062007 Powerful intelligence for long-running agents and coding20082009 - `"claude-opus-4-6"`20102011 Powerful intelligence for long-running agents and coding20122013 - `"claude-sonnet-4-6"`20142015 Best combination of speed and intelligence20162017 - `"claude-haiku-4-5"`20182019 Fastest model with near-frontier intelligence20202021 - `"claude-haiku-4-5-20251001"`20222023 Fastest model with near-frontier intelligence20242025 - `"claude-opus-4-5"`20262027 Powerful intelligence for long-running agents and coding20282029 - `"claude-opus-4-5-20251101"`20302031 Powerful intelligence for long-running agents and coding20322033 - `"claude-sonnet-4-5"`20342035 High-performance model for agents and coding20362037 - `"claude-sonnet-4-5-20250929"`20382039 High-performance model for agents and coding20402041 - `string`20422043 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`20442045 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.20462047 - `BetaManagedAgentsEffortLow object`20482049 Low effort. Favors latency over reasoning depth.20502051 - `type: "low"`20522053 - `BetaManagedAgentsEffortMedium object`20542055 Medium effort. Balances latency and reasoning depth.20562057 - `type: "medium"`20582059 - `BetaManagedAgentsEffortHigh object`20602061 High effort. Favors reasoning depth.20622063 - `type: "high"`20642065 - `BetaManagedAgentsEffortXhigh object`20662067 Extra-high effort. Not all models accept this level.20682069 - `type: "xhigh"`20702071 - `BetaManagedAgentsEffortMax object`20722073 Maximum effort. Favors reasoning depth over latency.20742075 - `type: "max"`20762077 - `inference_geo: optional string`20782079 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.20802081 - `speed: optional "standard" or "fast"`20822083 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.20842085 - `"standard"`20862087 - `"fast"`20882089 - `name: string`20902091 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`20922093 - `BetaManagedAgentsAnthropicSkill object`20942095 A resolved Anthropic-managed skill.20962097 - `type: "anthropic"`20982099 - `skill_id: string`21002101 - `version: string`21022103 - `BetaManagedAgentsCustomSkill object`21042105 A resolved user-created custom skill.21062107 - `type: "custom"`21082109 - `skill_id: string`21102111 - `version: string`21122113 - `system: string or null`21142115 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`21162117 - `BetaManagedAgentsAgentToolset20260401 object`21182119 - `type: "agent_toolset_20260401"`21202121 - `configs: array of BetaManagedAgentsAgentToolConfig`21222123 - `BetaManagedAgentsBashToolConfig object`21242125 Configuration for the bash tool.21262127 - `type: "bash"`21282129 - `enabled: boolean`21302131 - `name: "bash"`21322133 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`21342135 Permission policy for tool execution.21362137 - `BetaManagedAgentsAlwaysAllowPolicy object`21382139 Tool calls are automatically approved without user confirmation.21402141 - `type: "always_allow"`21422143 - `BetaManagedAgentsAlwaysAskPolicy object`21442145 Tool calls require user confirmation before execution.21462147 - `type: "always_ask"`21482149 - `BetaManagedAgentsAutoPolicy object`21502151 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.21522153 - `type: "auto"`21542155 - `BetaManagedAgentsEditToolConfig object`21562157 Configuration for the edit tool.21582159 - `type: "edit"`21602161 - `enabled: boolean`21622163 - `name: "edit"`21642165 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`21662167 Permission policy for tool execution.21682169 - `BetaManagedAgentsAlwaysAllowPolicy object`21702171 Tool calls are automatically approved without user confirmation.21722173 - `BetaManagedAgentsAlwaysAskPolicy object`21742175 Tool calls require user confirmation before execution.21762177 - `BetaManagedAgentsAutoPolicy object`21782179 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.21802181 - `BetaManagedAgentsReadToolConfig object`21822183 Configuration for the read tool.21842185 - `type: "read"`21862187 - `enabled: boolean`21882189 - `name: "read"`21902191 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`21922193 Permission policy for tool execution.21942195 - `BetaManagedAgentsAlwaysAllowPolicy object`21962197 Tool calls are automatically approved without user confirmation.21982199 - `BetaManagedAgentsAlwaysAskPolicy object`22002201 Tool calls require user confirmation before execution.22022203 - `BetaManagedAgentsAutoPolicy object`22042205 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.22062207 - `BetaManagedAgentsWriteToolConfig object`22082209 Configuration for the write tool.22102211 - `type: "write"`22122213 - `enabled: boolean`22142215 - `name: "write"`22162217 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`22182219 Permission policy for tool execution.22202221 - `BetaManagedAgentsAlwaysAllowPolicy object`22222223 Tool calls are automatically approved without user confirmation.22242225 - `BetaManagedAgentsAlwaysAskPolicy object`22262227 Tool calls require user confirmation before execution.22282229 - `BetaManagedAgentsAutoPolicy object`22302231 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.22322233 - `BetaManagedAgentsGlobToolConfig object`22342235 Configuration for the glob tool.22362237 - `type: "glob"`22382239 - `enabled: boolean`22402241 - `name: "glob"`22422243 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`22442245 Permission policy for tool execution.22462247 - `BetaManagedAgentsAlwaysAllowPolicy object`22482249 Tool calls are automatically approved without user confirmation.22502251 - `BetaManagedAgentsAlwaysAskPolicy object`22522253 Tool calls require user confirmation before execution.22542255 - `BetaManagedAgentsAutoPolicy object`22562257 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.22582259 - `BetaManagedAgentsGrepToolConfig object`22602261 Configuration for the grep tool.22622263 - `type: "grep"`22642265 - `enabled: boolean`22662267 - `name: "grep"`22682269 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`22702271 Permission policy for tool execution.22722273 - `BetaManagedAgentsAlwaysAllowPolicy object`22742275 Tool calls are automatically approved without user confirmation.22762277 - `BetaManagedAgentsAlwaysAskPolicy object`22782279 Tool calls require user confirmation before execution.22802281 - `BetaManagedAgentsAutoPolicy object`22822283 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.22842285 - `BetaManagedAgentsWebFetchToolConfig object`22862287 Configuration for the web_fetch tool.22882289 - `type: "web_fetch"`22902291 - `enabled: boolean`22922293 - `name: "web_fetch"`22942295 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`22962297 Permission policy for tool execution.22982299 - `BetaManagedAgentsAlwaysAllowPolicy object`23002301 Tool calls are automatically approved without user confirmation.23022303 - `BetaManagedAgentsAlwaysAskPolicy object`23042305 Tool calls require user confirmation before execution.23062307 - `BetaManagedAgentsAutoPolicy object`23082309 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.23102311 - `allowed_domains: optional array of string`23122313 - `blocked_domains: optional array of string`23142315 - `max_content_tokens: optional number or null`23162317 format: int3223182319 - `BetaManagedAgentsWebSearchToolConfig object`23202321 Configuration for the web_search tool.23222323 - `type: "web_search"`23242325 - `enabled: boolean`23262327 - `name: "web_search"`23282329 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`23302331 Permission policy for tool execution.23322333 - `BetaManagedAgentsAlwaysAllowPolicy object`23342335 Tool calls are automatically approved without user confirmation.23362337 - `BetaManagedAgentsAlwaysAskPolicy object`23382339 Tool calls require user confirmation before execution.23402341 - `BetaManagedAgentsAutoPolicy object`23422343 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.23442345 - `allowed_domains: optional array of string`23462347 - `blocked_domains: optional array of string`23482349 - `user_location: optional BetaManagedAgentsUserLocation or null`23502351 Approximate user location for search result localization.23522353 - `type: "approximate"`23542355 Location precision. Only "approximate" is supported.23562357 - `city: optional string or null`23582359 City name.23602361 minLength: 1, maxLength: 25523622363 - `country: optional string or null`23642365 Two-letter ISO 3166-1 country code, uppercase.23662367 - `region: optional string or null`23682369 Region or state name.23702371 minLength: 1, maxLength: 25523722373 - `timezone: optional string or null`23742375 IANA timezone identifier, e.g. "America/Los_Angeles".23762377 minLength: 1, maxLength: 25523782379 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`23802381 Resolved default configuration for agent tools.23822383 - `enabled: boolean`23842385 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`23862387 Permission policy for tool execution.23882389 - `BetaManagedAgentsAlwaysAllowPolicy object`23902391 Tool calls are automatically approved without user confirmation.23922393 - `BetaManagedAgentsAlwaysAskPolicy object`23942395 Tool calls require user confirmation before execution.23962397 - `BetaManagedAgentsAutoPolicy object`23982399 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.24002401 - `BetaManagedAgentsMCPToolset object`24022403 - `type: "mcp_toolset"`24042405 - `configs: array of BetaManagedAgentsMCPToolConfig`24062407 - `enabled: boolean`24082409 - `name: string`24102411 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`24122413 Permission policy for tool execution.24142415 - `BetaManagedAgentsAlwaysAllowPolicy object`24162417 Tool calls are automatically approved without user confirmation.24182419 - `BetaManagedAgentsAlwaysAskPolicy object`24202421 Tool calls require user confirmation before execution.24222423 - `BetaManagedAgentsAutoPolicy object`24242425 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.24262427 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`24282429 Resolved default configuration for all tools from an MCP server.24302431 - `enabled: boolean`24322433 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`24342435 Permission policy for tool execution.24362437 - `BetaManagedAgentsAlwaysAllowPolicy object`24382439 Tool calls are automatically approved without user confirmation.24402441 - `BetaManagedAgentsAlwaysAskPolicy object`24422443 Tool calls require user confirmation before execution.24442445 - `BetaManagedAgentsAutoPolicy object`24462447 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.24482449 - `mcp_server_name: string`24502451 - `BetaManagedAgentsCustomTool object`24522453 A custom tool as returned in API responses.24542455 - `type: "custom"`24562457 - `description: string`24582459 - `input_schema: BetaManagedAgentsCustomToolInputSchema`24602461 JSON Schema for custom tool input parameters.24622463 - `type: "object"`24642465 - `properties: optional map[unknown] or null`24662467 - `required: optional array of string or null`24682469 - `name: string`24702471 - `version: number`24722473 format: int3224742475 - `BetaManagedAgentsAdvisor object`24762477 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.24782479 - `type: "advisor"`24802481 - `model: string`24822483 The advisor model id.24842485 - `archived_at: string or null`24862487 A timestamp in RFC 3339 format24882489 format: date-time24902491 - `created_at: string`24922493 A timestamp in RFC 3339 format24942495 format: date-time24962497 - `parent_thread_id: string or null`24982499 Parent thread that spawned this thread. Null for the primary thread.25002501 - `session_id: string`25022503 The session this thread belongs to.25042505 - `stats: BetaManagedAgentsSessionThreadStats or null`25062507 Timing statistics for a session thread.25082509 - `active_seconds: optional number`25102511 Cumulative time in seconds the thread spent actively running. Excludes idle time.25122513 format: double25142515 - `duration_seconds: optional number`25162517 Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.25182519 format: double25202521 - `startup_seconds: optional number`25222523 Time in seconds for the thread to begin running. Zero for child threads, which start immediately.25242525 format: double25262527 - `status: BetaManagedAgentsSessionThreadStatus`25282529 SessionThreadStatus enum25302531 - `"running"`25322533 - `"idle"`25342535 - `"rescheduling"`25362537 - `"terminated"`25382539 - `updated_at: string`25402541 A timestamp in RFC 3339 format25422543 format: date-time25442545 - `usage: BetaManagedAgentsSessionThreadUsage or null`25462547 Cumulative token usage for a session thread across all turns.25482549 - `active_seconds: optional number`25502551 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.25522553 format: double25542555 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`25562557 Prompt-cache creation token usage broken down by cache lifetime.25582559 - `ephemeral_1h_input_tokens: optional number`25602561 Tokens used to create 1-hour ephemeral cache entries.25622563 format: int3225642565 - `ephemeral_5m_input_tokens: optional number`25662567 Tokens used to create 5-minute ephemeral cache entries.25682569 format: int3225702571 - `cache_read_input_tokens: optional number`25722573 Total tokens read from prompt cache.25742575 format: int3225762577 - `input_tokens: optional number`25782579 Total input tokens consumed across all turns.25802581 format: int3225822583 - `list_cost: optional BetaMonetaryAmount or null`25842585 A monetary amount in a specific currency.25862587 - `amount: string`25882589 Amount in minor units of the currency, as an integer decimal string with no leading zeros: "2500" is $25.00 and "50" is fifty cents. A string rather than a number so no float rounding is ever applied.25902591 - `currency: BetaCurrency`25922593 Uppercase ISO-4217 currency code. `USD` is the only currency currently supported; the accepted set is closed and grows only when a new currency is priced.25942595 - `output_tokens: optional number`25962597 Total output tokens generated across all turns.25982599 format: int3226002601 - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`26022603 Cumulative count of server-executed tool invocations, broken down by tool.26042605 - `web_fetch_requests: optional number`26062607 Number of server-executed web fetch requests.26082609 format: int3226102611 - `web_search_requests: optional number`26122613 Number of server-executed web search requests.26142615 format: int3226162617### Example26182619```bash2620curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID/archive \2621 -X POST \2622 -H 'anthropic-version: 2023-06-01' \2623 -H 'anthropic-beta: managed-agents-2026-04-01' \2624 -H "X-Api-Key: $ANTHROPIC_API_KEY"2625```26262627#### Response (200)26282629```json2630{2631 "id": "sthr_011CZkZVWa6oIjw0rgXZpnBt",2632 "agent": {2633 "id": "agent_011CZkYqphY8vELVzwCUpqiQ",2634 "description": "A focused research subagent.",2635 "mcp_servers": [2636 {2637 "name": "example-mcp",2638 "type": "url",2639 "url": "https://example-server.modelcontextprotocol.io/sse"2640 }2641 ],2642 "model": {2643 "id": "claude-opus-5",2644 "effort": {2645 "type": "low"2646 },2647 "inference_geo": "inference_geo",2648 "speed": "standard"2649 },2650 "name": "Researcher",2651 "skills": [2652 {2653 "skill_id": "xlsx",2654 "type": "anthropic",2655 "version": "1"2656 }2657 ],2658 "system": "You are a research subagent that gathers and summarises sources for the coordinating agent.",2659 "tools": [2660 {2661 "configs": [2662 {2663 "enabled": true,2664 "name": "bash",2665 "permission_policy": {2666 "type": "always_allow"2667 },2668 "type": "bash"2669 }2670 ],2671 "default_config": {2672 "enabled": true,2673 "permission_policy": {2674 "type": "always_ask"2675 }2676 },2677 "type": "agent_toolset_20260401"2678 }2679 ],2680 "type": "agent",2681 "version": 12682 },2683 "archived_at": null,2684 "created_at": "2026-03-15T10:00:00Z",2685 "parent_thread_id": null,2686 "session_id": "sesn_011CZkZAtmR3yMPDzynEDxu7",2687 "stats": {2688 "active_seconds": 0,2689 "duration_seconds": 0,2690 "startup_seconds": 02691 },2692 "status": "idle",2693 "type": "session_thread",2694 "updated_at": "2026-03-15T10:00:00Z",2695 "usage": {2696 "active_seconds": 0,2697 "cache_creation": {2698 "ephemeral_1h_input_tokens": 0,2699 "ephemeral_5m_input_tokens": 02700 },2701 "cache_read_input_tokens": 0,2702 "input_tokens": 0,2703 "list_cost": {2704 "amount": "2500",2705 "currency": "USD"2706 },2707 "output_tokens": 0,2708 "server_tool_use": {2709 "web_fetch_requests": 0,2710 "web_search_requests": 32711 }2712 }2713}2714```27152716## Domain types27172718### Beta Managed Agents Session Thread27192720- `BetaManagedAgentsSessionThread object`27212722 An execution thread within a `session`. Each session has one primary thread plus zero or more child threads spawned by the coordinator.27232724 - `type: "session_thread"`27252726 - `id: string`27272728 Unique identifier for this thread.27292730 - `agent: BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`27312732 The resolved agent a session thread runs: a saved-agent snapshot, the platform advisor entry, or an inline-defined (ephemeral) agent snapshot.27332734 - `BetaManagedAgentsSessionThreadAgent object`27352736 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`.27372738 - `type: "agent"`27392740 - `id: string`27412742 - `description: string or null`27432744 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`27452746 - `type: "url"`27472748 - `name: string`27492750 - `url: string`27512752 - `model: BetaManagedAgentsModelConfig`27532754 Model identifier and configuration.27552756 - `id: BetaManagedAgentsModel`27572758 The model that will power your agent.27592760 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.27612762 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`27632764 The model that will power your agent.27652766 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.27672768 - `"claude-fable-5-1"`27692770 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows27712772 - `"claude-sonnet-5"`27732774 High-performance model for coding and agents27752776 - `"claude-fable-5"`27772778 Next generation of intelligence for the hardest knowledge work and coding problems27792780 - `"claude-opus-5"`27812782 Powerful intelligence for long-running agents and coding27832784 - `"claude-opus-4-8"`27852786 Powerful intelligence for long-running agents and coding27872788 - `"claude-opus-4-7"`27892790 Powerful intelligence for long-running agents and coding27912792 - `"claude-opus-4-6"`27932794 Powerful intelligence for long-running agents and coding27952796 - `"claude-sonnet-4-6"`27972798 Best combination of speed and intelligence27992800 - `"claude-haiku-4-5"`28012802 Fastest model with near-frontier intelligence28032804 - `"claude-haiku-4-5-20251001"`28052806 Fastest model with near-frontier intelligence28072808 - `"claude-opus-4-5"`28092810 Powerful intelligence for long-running agents and coding28112812 - `"claude-opus-4-5-20251101"`28132814 Powerful intelligence for long-running agents and coding28152816 - `"claude-sonnet-4-5"`28172818 High-performance model for agents and coding28192820 - `"claude-sonnet-4-5-20250929"`28212822 High-performance model for agents and coding28232824 - `string`28252826 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`28272828 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.28292830 - `BetaManagedAgentsEffortLow object`28312832 Low effort. Favors latency over reasoning depth.28332834 - `type: "low"`28352836 - `BetaManagedAgentsEffortMedium object`28372838 Medium effort. Balances latency and reasoning depth.28392840 - `type: "medium"`28412842 - `BetaManagedAgentsEffortHigh object`28432844 High effort. Favors reasoning depth.28452846 - `type: "high"`28472848 - `BetaManagedAgentsEffortXhigh object`28492850 Extra-high effort. Not all models accept this level.28512852 - `type: "xhigh"`28532854 - `BetaManagedAgentsEffortMax object`28552856 Maximum effort. Favors reasoning depth over latency.28572858 - `type: "max"`28592860 - `inference_geo: optional string`28612862 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.28632864 - `speed: optional "standard" or "fast"`28652866 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.28672868 - `"standard"`28692870 - `"fast"`28712872 - `name: string`28732874 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`28752876 - `BetaManagedAgentsAnthropicSkill object`28772878 A resolved Anthropic-managed skill.28792880 - `type: "anthropic"`28812882 - `skill_id: string`28832884 - `version: string`28852886 - `BetaManagedAgentsCustomSkill object`28872888 A resolved user-created custom skill.28892890 - `type: "custom"`28912892 - `skill_id: string`28932894 - `version: string`28952896 - `system: string or null`28972898 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`28992900 - `BetaManagedAgentsAgentToolset20260401 object`29012902 - `type: "agent_toolset_20260401"`29032904 - `configs: array of BetaManagedAgentsAgentToolConfig`29052906 - `BetaManagedAgentsBashToolConfig object`29072908 Configuration for the bash tool.29092910 - `type: "bash"`29112912 - `enabled: boolean`29132914 - `name: "bash"`29152916 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`29172918 Permission policy for tool execution.29192920 - `BetaManagedAgentsAlwaysAllowPolicy object`29212922 Tool calls are automatically approved without user confirmation.29232924 - `type: "always_allow"`29252926 - `BetaManagedAgentsAlwaysAskPolicy object`29272928 Tool calls require user confirmation before execution.29292930 - `type: "always_ask"`29312932 - `BetaManagedAgentsAutoPolicy object`29332934 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.29352936 - `type: "auto"`29372938 - `BetaManagedAgentsEditToolConfig object`29392940 Configuration for the edit tool.29412942 - `type: "edit"`29432944 - `enabled: boolean`29452946 - `name: "edit"`29472948 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`29492950 Permission policy for tool execution.29512952 - `BetaManagedAgentsAlwaysAllowPolicy object`29532954 Tool calls are automatically approved without user confirmation.29552956 - `BetaManagedAgentsAlwaysAskPolicy object`29572958 Tool calls require user confirmation before execution.29592960 - `BetaManagedAgentsAutoPolicy object`29612962 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.29632964 - `BetaManagedAgentsReadToolConfig object`29652966 Configuration for the read tool.29672968 - `type: "read"`29692970 - `enabled: boolean`29712972 - `name: "read"`29732974 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`29752976 Permission policy for tool execution.29772978 - `BetaManagedAgentsAlwaysAllowPolicy object`29792980 Tool calls are automatically approved without user confirmation.29812982 - `BetaManagedAgentsAlwaysAskPolicy object`29832984 Tool calls require user confirmation before execution.29852986 - `BetaManagedAgentsAutoPolicy object`29872988 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.29892990 - `BetaManagedAgentsWriteToolConfig object`29912992 Configuration for the write tool.29932994 - `type: "write"`29952996 - `enabled: boolean`29972998 - `name: "write"`29993000 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`30013002 Permission policy for tool execution.30033004 - `BetaManagedAgentsAlwaysAllowPolicy object`30053006 Tool calls are automatically approved without user confirmation.30073008 - `BetaManagedAgentsAlwaysAskPolicy object`30093010 Tool calls require user confirmation before execution.30113012 - `BetaManagedAgentsAutoPolicy object`30133014 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.30153016 - `BetaManagedAgentsGlobToolConfig object`30173018 Configuration for the glob tool.30193020 - `type: "glob"`30213022 - `enabled: boolean`30233024 - `name: "glob"`30253026 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`30273028 Permission policy for tool execution.30293030 - `BetaManagedAgentsAlwaysAllowPolicy object`30313032 Tool calls are automatically approved without user confirmation.30333034 - `BetaManagedAgentsAlwaysAskPolicy object`30353036 Tool calls require user confirmation before execution.30373038 - `BetaManagedAgentsAutoPolicy object`30393040 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.30413042 - `BetaManagedAgentsGrepToolConfig object`30433044 Configuration for the grep tool.30453046 - `type: "grep"`30473048 - `enabled: boolean`30493050 - `name: "grep"`30513052 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`30533054 Permission policy for tool execution.30553056 - `BetaManagedAgentsAlwaysAllowPolicy object`30573058 Tool calls are automatically approved without user confirmation.30593060 - `BetaManagedAgentsAlwaysAskPolicy object`30613062 Tool calls require user confirmation before execution.30633064 - `BetaManagedAgentsAutoPolicy object`30653066 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.30673068 - `BetaManagedAgentsWebFetchToolConfig object`30693070 Configuration for the web_fetch tool.30713072 - `type: "web_fetch"`30733074 - `enabled: boolean`30753076 - `name: "web_fetch"`30773078 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`30793080 Permission policy for tool execution.30813082 - `BetaManagedAgentsAlwaysAllowPolicy object`30833084 Tool calls are automatically approved without user confirmation.30853086 - `BetaManagedAgentsAlwaysAskPolicy object`30873088 Tool calls require user confirmation before execution.30893090 - `BetaManagedAgentsAutoPolicy object`30913092 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.30933094 - `allowed_domains: optional array of string`30953096 - `blocked_domains: optional array of string`30973098 - `max_content_tokens: optional number or null`30993100 format: int3231013102 - `BetaManagedAgentsWebSearchToolConfig object`31033104 Configuration for the web_search tool.31053106 - `type: "web_search"`31073108 - `enabled: boolean`31093110 - `name: "web_search"`31113112 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`31133114 Permission policy for tool execution.31153116 - `BetaManagedAgentsAlwaysAllowPolicy object`31173118 Tool calls are automatically approved without user confirmation.31193120 - `BetaManagedAgentsAlwaysAskPolicy object`31213122 Tool calls require user confirmation before execution.31233124 - `BetaManagedAgentsAutoPolicy object`31253126 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.31273128 - `allowed_domains: optional array of string`31293130 - `blocked_domains: optional array of string`31313132 - `user_location: optional BetaManagedAgentsUserLocation or null`31333134 Approximate user location for search result localization.31353136 - `type: "approximate"`31373138 Location precision. Only "approximate" is supported.31393140 - `city: optional string or null`31413142 City name.31433144 minLength: 1, maxLength: 25531453146 - `country: optional string or null`31473148 Two-letter ISO 3166-1 country code, uppercase.31493150 - `region: optional string or null`31513152 Region or state name.31533154 minLength: 1, maxLength: 25531553156 - `timezone: optional string or null`31573158 IANA timezone identifier, e.g. "America/Los_Angeles".31593160 minLength: 1, maxLength: 25531613162 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`31633164 Resolved default configuration for agent tools.31653166 - `enabled: boolean`31673168 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`31693170 Permission policy for tool execution.31713172 - `BetaManagedAgentsAlwaysAllowPolicy object`31733174 Tool calls are automatically approved without user confirmation.31753176 - `BetaManagedAgentsAlwaysAskPolicy object`31773178 Tool calls require user confirmation before execution.31793180 - `BetaManagedAgentsAutoPolicy object`31813182 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.31833184 - `BetaManagedAgentsMCPToolset object`31853186 - `type: "mcp_toolset"`31873188 - `configs: array of BetaManagedAgentsMCPToolConfig`31893190 - `enabled: boolean`31913192 - `name: string`31933194 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`31953196 Permission policy for tool execution.31973198 - `BetaManagedAgentsAlwaysAllowPolicy object`31993200 Tool calls are automatically approved without user confirmation.32013202 - `BetaManagedAgentsAlwaysAskPolicy object`32033204 Tool calls require user confirmation before execution.32053206 - `BetaManagedAgentsAutoPolicy object`32073208 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.32093210 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`32113212 Resolved default configuration for all tools from an MCP server.32133214 - `enabled: boolean`32153216 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`32173218 Permission policy for tool execution.32193220 - `BetaManagedAgentsAlwaysAllowPolicy object`32213222 Tool calls are automatically approved without user confirmation.32233224 - `BetaManagedAgentsAlwaysAskPolicy object`32253226 Tool calls require user confirmation before execution.32273228 - `BetaManagedAgentsAutoPolicy object`32293230 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.32313232 - `mcp_server_name: string`32333234 - `BetaManagedAgentsCustomTool object`32353236 A custom tool as returned in API responses.32373238 - `type: "custom"`32393240 - `description: string`32413242 - `input_schema: BetaManagedAgentsCustomToolInputSchema`32433244 JSON Schema for custom tool input parameters.32453246 - `type: "object"`32473248 - `properties: optional map[unknown] or null`32493250 - `required: optional array of string or null`32513252 - `name: string`32533254 - `version: number`32553256 format: int3232573258 - `BetaManagedAgentsAdvisor object`32593260 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.32613262 - `type: "advisor"`32633264 - `model: string`32653266 The advisor model id.32673268 - `archived_at: string or null`32693270 A timestamp in RFC 3339 format32713272 format: date-time32733274 - `created_at: string`32753276 A timestamp in RFC 3339 format32773278 format: date-time32793280 - `parent_thread_id: string or null`32813282 Parent thread that spawned this thread. Null for the primary thread.32833284 - `session_id: string`32853286 The session this thread belongs to.32873288 - `stats: BetaManagedAgentsSessionThreadStats or null`32893290 Timing statistics for a session thread.32913292 - `active_seconds: optional number`32933294 Cumulative time in seconds the thread spent actively running. Excludes idle time.32953296 format: double32973298 - `duration_seconds: optional number`32993300 Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.33013302 format: double33033304 - `startup_seconds: optional number`33053306 Time in seconds for the thread to begin running. Zero for child threads, which start immediately.33073308 format: double33093310 - `status: BetaManagedAgentsSessionThreadStatus`33113312 SessionThreadStatus enum33133314 - `"running"`33153316 - `"idle"`33173318 - `"rescheduling"`33193320 - `"terminated"`33213322 - `updated_at: string`33233324 A timestamp in RFC 3339 format33253326 format: date-time33273328 - `usage: BetaManagedAgentsSessionThreadUsage or null`33293330 Cumulative token usage for a session thread across all turns.33313332 - `active_seconds: optional number`33333334 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.33353336 format: double33373338 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`33393340 Prompt-cache creation token usage broken down by cache lifetime.33413342 - `ephemeral_1h_input_tokens: optional number`33433344 Tokens used to create 1-hour ephemeral cache entries.33453346 format: int3233473348 - `ephemeral_5m_input_tokens: optional number`33493350 Tokens used to create 5-minute ephemeral cache entries.33513352 format: int3233533354 - `cache_read_input_tokens: optional number`33553356 Total tokens read from prompt cache.33573358 format: int3233593360 - `input_tokens: optional number`33613362 Total input tokens consumed across all turns.33633364 format: int3233653366 - `list_cost: optional BetaMonetaryAmount or null`33673368 A monetary amount in a specific currency.33693370 - `amount: string`33713372 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.33733374 - `currency: BetaCurrency`33753376 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.33773378 - `output_tokens: optional number`33793380 Total output tokens generated across all turns.33813382 format: int3233833384 - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`33853386 Cumulative count of server-executed tool invocations, broken down by tool.33873388 - `web_fetch_requests: optional number`33893390 Number of server-executed web fetch requests.33913392 format: int3233933394 - `web_search_requests: optional number`33953396 Number of server-executed web search requests.33973398 format: int3233993400### Beta Managed Agents Session Thread Stats34013402- `BetaManagedAgentsSessionThreadStats object`34033404 Timing statistics for a session thread.34053406 - `active_seconds: optional number`34073408 Cumulative time in seconds the thread spent actively running. Excludes idle time.34093410 format: double34113412 - `duration_seconds: optional number`34133414 Elapsed time since thread creation in seconds. For archived threads, frozen at the final update.34153416 format: double34173418 - `startup_seconds: optional number`34193420 Time in seconds for the thread to begin running. Zero for child threads, which start immediately.34213422 format: double34233424### Beta Managed Agents Session Thread Status34253426- `BetaManagedAgentsSessionThreadStatus = "running" or "idle" or "rescheduling" or "terminated"`34273428 SessionThreadStatus enum34293430 - `"running"`34313432 - `"idle"`34333434 - `"rescheduling"`34353436 - `"terminated"`34373438### Beta Managed Agents Session Thread Usage34393440- `BetaManagedAgentsSessionThreadUsage object`34413442 Cumulative token usage for a session thread across all turns.34433444 - `active_seconds: optional number`34453446 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.34473448 format: double34493450 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`34513452 Prompt-cache creation token usage broken down by cache lifetime.34533454 - `ephemeral_1h_input_tokens: optional number`34553456 Tokens used to create 1-hour ephemeral cache entries.34573458 format: int3234593460 - `ephemeral_5m_input_tokens: optional number`34613462 Tokens used to create 5-minute ephemeral cache entries.34633464 format: int3234653466 - `cache_read_input_tokens: optional number`34673468 Total tokens read from prompt cache.34693470 format: int3234713472 - `input_tokens: optional number`34733474 Total input tokens consumed across all turns.34753476 format: int3234773478 - `list_cost: optional BetaMonetaryAmount or null`34793480 A monetary amount in a specific currency.34813482 - `amount: string`34833484 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.34853486 - `currency: BetaCurrency`34873488 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.34893490 - `output_tokens: optional number`34913492 Total output tokens generated across all turns.34933494 format: int3234953496 - `server_tool_use: optional BetaManagedAgentsServerToolUsage or null`34973498 Cumulative count of server-executed tool invocations, broken down by tool.34993500 - `web_fetch_requests: optional number`35013502 Number of server-executed web fetch requests.35033504 format: int3235053506 - `web_search_requests: optional number`35073508 Number of server-executed web search requests.35093510 format: int3235113512### Beta Managed Agents Stream Session Thread Events35133514- `BetaManagedAgentsStreamSessionThreadEvents = BetaManagedAgentsUserMessageEvent or BetaManagedAgentsUserInterruptEvent or BetaManagedAgentsUserToolConfirmationEvent or 34 more`35153516 Server-sent event in a single thread's stream.35173518 - `BetaManagedAgentsUserMessageEvent object`35193520 A user message event in the session conversation.35213522 - `type: "user.message"`35233524 - `id: string`35253526 Unique identifier for this event.35273528 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`35293530 Array of content blocks comprising the user message.35313532 - `BetaManagedAgentsTextBlock object`35333534 Regular text content.35353536 - `type: "text"`35373538 - `text: string`35393540 The text content.35413542 minLength: 135433544 - `BetaManagedAgentsImageBlock object`35453546 Image content specified directly as base64 data or as a reference via a URL.35473548 - `type: "image"`35493550 - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`35513552 Union type for image source variants.35533554 - `BetaManagedAgentsBase64ImageSource object`35553556 Base64-encoded image data.35573558 - `type: "base64"`35593560 - `data: string`35613562 Base64-encoded image data.35633564 minLength: 135653566 - `media_type: string`35673568 MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").35693570 minLength: 135713572 - `BetaManagedAgentsURLImageSource object`35733574 Image referenced by URL.35753576 - `type: "url"`35773578 - `url: string`35793580 URL of the image to fetch.35813582 minLength: 135833584 - `BetaManagedAgentsFileImageSource object`35853586 Image referenced by file ID.35873588 - `type: "file"`35893590 - `file_id: string`35913592 ID of a previously uploaded file.35933594 minLength: 135953596 - `BetaManagedAgentsDocumentBlock object`35973598 Document content, either specified directly as base64 data, as text, or as a reference via a URL.35993600 - `type: "document"`36013602 - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`36033604 Union type for document source variants.36053606 - `BetaManagedAgentsBase64DocumentSource object`36073608 Base64-encoded document data.36093610 - `type: "base64"`36113612 - `data: string`36133614 Base64-encoded document data.36153616 minLength: 136173618 - `media_type: string`36193620 MIME type of the document (e.g., "application/pdf").36213622 minLength: 136233624 - `BetaManagedAgentsPlainTextDocumentSource object`36253626 Plain text document content.36273628 - `type: "text"`36293630 - `data: string`36313632 The plain text content.36333634 minLength: 136353636 - `media_type: "text/plain"`36373638 MIME type of the text content. Must be "text/plain".36393640 - `BetaManagedAgentsURLDocumentSource object`36413642 Document referenced by URL.36433644 - `type: "url"`36453646 - `url: string`36473648 URL of the document to fetch.36493650 minLength: 136513652 - `BetaManagedAgentsFileDocumentSource object`36533654 Document referenced by file ID.36553656 - `type: "file"`36573658 - `file_id: string`36593660 ID of a previously uploaded file.36613662 minLength: 136633664 - `context: optional string or null`36653666 Additional context about the document for the model.36673668 - `title: optional string or null`36693670 The title of the document.36713672 - `BetaManagedAgentsRedactedBlock object`36733674 Placeholder for content withheld by Anthropic model policy.36753676 - `type: "redacted"`36773678 - `processed_at: optional string or null`36793680 A timestamp in RFC 3339 format36813682 format: date-time36833684 - `BetaManagedAgentsUserInterruptEvent object`36853686 An interrupt event that pauses agent execution and returns control to the user.36873688 - `type: "user.interrupt"`36893690 - `id: string`36913692 Unique identifier for this event.36933694 - `processed_at: optional string or null`36953696 A timestamp in RFC 3339 format36973698 format: date-time36993700 - `session_thread_id: optional string or null`37013702 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.37033704 - `BetaManagedAgentsUserToolConfirmationEvent object`37053706 A tool confirmation event that approves or denies a pending tool execution.37073708 - `type: "user.tool_confirmation"`37093710 - `id: string`37113712 Unique identifier for this event.37133714 - `result: "allow" or "deny"`37153716 UserToolConfirmationResult enum37173718 - `"allow"`37193720 - `"deny"`37213722 - `tool_use_id: string`37233724 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.37253726 - `deny_message: optional string or null`37273728 Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.37293730 maxLength: 1000037313732 - `processed_at: optional string or null`37333734 A timestamp in RFC 3339 format37353736 format: date-time37373738 - `session_thread_id: optional string or null`37393740 Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.37413742 - `BetaManagedAgentsUserCustomToolResultEvent object`37433744 Event sent by the client providing the result of a custom tool execution.37453746 - `type: "user.custom_tool_result"`37473748 - `id: string`37493750 Unique identifier for this event.37513752 - `custom_tool_use_id: string`37533754 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.37553756 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`37573758 The result content returned by the tool.37593760 - `BetaManagedAgentsTextBlock object`37613762 Regular text content.37633764 - `BetaManagedAgentsImageBlock object`37653766 Image content specified directly as base64 data or as a reference via a URL.37673768 - `BetaManagedAgentsDocumentBlock object`37693770 Document content, either specified directly as base64 data, as text, or as a reference via a URL.37713772 - `BetaManagedAgentsSearchResultBlock object`37733774 A block containing a web search result.37753776 - `type: "search_result"`37773778 - `citations: BetaManagedAgentsSearchResultCitations`37793780 Citation settings for a search result.37813782 - `enabled: boolean`37833784 Whether citations are enabled for this search result.37853786 - `content: array of BetaManagedAgentsSearchResultContent`37873788 Array of text content blocks from the search result.37893790 - `type: "text"`37913792 - `text: string`37933794 The text content.37953796 minLength: 137973798 - `source: string`37993800 The URL source of the search result.38013802 minLength: 138033804 - `title: string`38053806 The title of the search result.38073808 minLength: 138093810 - `is_error: optional boolean or null`38113812 Whether the tool execution resulted in an error.38133814 - `processed_at: optional string or null`38153816 A timestamp in RFC 3339 format38173818 format: date-time38193820 - `session_thread_id: optional string or null`38213822 Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.38233824 - `BetaManagedAgentsAgentCustomToolUseEvent object`38253826 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.38273828 - `type: "agent.custom_tool_use"`38293830 - `id: string`38313832 Unique identifier for this event.38333834 - `input: map[unknown]`38353836 Input parameters for the tool call.38373838 - `name: string`38393840 Name of the custom tool being called.38413842 - `processed_at: string`38433844 A timestamp in RFC 3339 format38453846 format: date-time38473848 - `session_thread_id: optional string or null`38493850 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.38513852 - `BetaManagedAgentsAgentMessageEvent object`38533854 An agent response event in the session conversation.38553856 - `type: "agent.message"`38573858 - `id: string`38593860 Unique identifier for this event.38613862 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`38633864 Array of text blocks comprising the agent response.38653866 - `BetaManagedAgentsTextBlock object`38673868 Regular text content.38693870 - `BetaManagedAgentsRedactedBlock object`38713872 Placeholder for content withheld by Anthropic model policy.38733874 - `processed_at: string`38753876 A timestamp in RFC 3339 format38773878 format: date-time38793880 - `BetaManagedAgentsAgentThinkingEvent object`38813882 Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.38833884 - `type: "agent.thinking"`38853886 - `id: string`38873888 Unique identifier for this event.38893890 - `processed_at: string`38913892 A timestamp in RFC 3339 format38933894 format: date-time38953896 - `BetaManagedAgentsAgentMCPToolUseEvent object`38973898 Event emitted when the agent invokes a tool provided by an MCP server.38993900 - `type: "agent.mcp_tool_use"`39013902 - `id: string`39033904 Unique identifier for this event.39053906 - `input: map[unknown]`39073908 Input parameters for the tool call.39093910 - `mcp_server_name: string`39113912 Name of the MCP server providing the tool.39133914 - `name: string`39153916 Name of the MCP tool being used.39173918 - `processed_at: string`39193920 A timestamp in RFC 3339 format39213922 format: date-time39233924 - `evaluated_permission: optional "allow" or "ask" or "deny"`39253926 AgentEvaluatedPermission enum39273928 - `"allow"`39293930 - `"ask"`39313932 - `"deny"`39333934 - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`39353936 Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.39373938 - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`39393940 The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".39413942 - `type: "always_allow"`39433944 - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`39453946 The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".39473948 - `type: "always_ask"`39493950 - `BetaManagedAgentsAgentToolEvaluationAuto object`39513952 The resolved permission_policy was auto: the server judged this invocation individually.39533954 - `type: "auto"`39553956 - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`39573958 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.39593960 - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`39613962 The server judged the invocation safe to execute without client approval.39633964 - `type: "allow"`39653966 - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`39673968 The server reached no judgement; the invocation is held for client approval.39693970 - `type: "ask"`39713972 - `reason_code: string`39733974 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.39753976 maxLength: 6439773978 - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`39793980 The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.39813982 - `type: "deny"`39833984 - `reason_code: string`39853986 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.39873988 maxLength: 6439893990 - `session_thread_id: optional string or null`39913992 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.39933994 - `BetaManagedAgentsAgentMCPToolResultEvent object`39953996 Event representing the result of an MCP tool execution.39973998 - `type: "agent.mcp_tool_result"`39994000 - `id: string`40014002 Unique identifier for this event.40034004 - `mcp_tool_use_id: string`40054006 The id of the `agent.mcp_tool_use` event this result corresponds to.40074008 - `processed_at: string`40094010 A timestamp in RFC 3339 format40114012 format: date-time40134014 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`40154016 The result content returned by the tool.40174018 - `BetaManagedAgentsTextBlock object`40194020 Regular text content.40214022 - `BetaManagedAgentsImageBlock object`40234024 Image content specified directly as base64 data or as a reference via a URL.40254026 - `BetaManagedAgentsDocumentBlock object`40274028 Document content, either specified directly as base64 data, as text, or as a reference via a URL.40294030 - `BetaManagedAgentsSearchResultBlock object`40314032 A block containing a web search result.40334034 - `is_error: optional boolean or null`40354036 Whether the tool execution resulted in an error.40374038 - `BetaManagedAgentsAgentToolUseEvent object`40394040 Event emitted when the agent invokes a built-in agent tool.40414042 - `type: "agent.tool_use"`40434044 - `id: string`40454046 Unique identifier for this event.40474048 - `input: map[unknown]`40494050 Input parameters for the tool call.40514052 - `name: string`40534054 Name of the agent tool being used.40554056 - `processed_at: string`40574058 A timestamp in RFC 3339 format40594060 format: date-time40614062 - `evaluated_permission: optional "allow" or "ask" or "deny"`40634064 AgentEvaluatedPermission enum40654066 - `"allow"`40674068 - `"ask"`40694070 - `"deny"`40714072 - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`40734074 Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.40754076 - `session_thread_id: optional string or null`40774078 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.40794080 - `BetaManagedAgentsAgentToolResultEvent object`40814082 Event representing the result of an agent tool execution.40834084 - `type: "agent.tool_result"`40854086 - `id: string`40874088 Unique identifier for this event.40894090 - `processed_at: string`40914092 A timestamp in RFC 3339 format40934094 format: date-time40954096 - `tool_use_id: string`40974098 The id of the `agent.tool_use` event this result corresponds to.40994100 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`41014102 The result content returned by the tool.41034104 - `BetaManagedAgentsTextBlock object`41054106 Regular text content.41074108 - `BetaManagedAgentsImageBlock object`41094110 Image content specified directly as base64 data or as a reference via a URL.41114112 - `BetaManagedAgentsDocumentBlock object`41134114 Document content, either specified directly as base64 data, as text, or as a reference via a URL.41154116 - `BetaManagedAgentsSearchResultBlock object`41174118 A block containing a web search result.41194120 - `is_error: optional boolean or null`41214122 Whether the tool execution resulted in an error.41234124 - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`41254126 Delivery event written to the target thread's input stream when an agent-to-agent message arrives.41274128 - `type: "agent.thread_message_received"`41294130 - `id: string`41314132 Unique identifier for this event.41334134 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`41354136 Message content blocks.41374138 - `BetaManagedAgentsTextBlock object`41394140 Regular text content.41414142 - `BetaManagedAgentsImageBlock object`41434144 Image content specified directly as base64 data or as a reference via a URL.41454146 - `BetaManagedAgentsDocumentBlock object`41474148 Document content, either specified directly as base64 data, as text, or as a reference via a URL.41494150 - `BetaManagedAgentsRedactedBlock object`41514152 Placeholder for content withheld by Anthropic model policy.41534154 - `from_session_thread_id: string`41554156 Public `sthr_` ID of the thread that sent the message.41574158 - `processed_at: string`41594160 A timestamp in RFC 3339 format41614162 format: date-time41634164 - `from_agent_name: optional string or null`41654166 Name of the callable agent this message came from. Absent when received from the primary agent.41674168 - `BetaManagedAgentsAgentThreadMessageSentEvent object`41694170 Observability event emitted to the sender's output stream when an agent-to-agent message is sent.41714172 - `type: "agent.thread_message_sent"`41734174 - `id: string`41754176 Unique identifier for this event.41774178 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`41794180 Message content blocks.41814182 - `BetaManagedAgentsTextBlock object`41834184 Regular text content.41854186 - `BetaManagedAgentsImageBlock object`41874188 Image content specified directly as base64 data or as a reference via a URL.41894190 - `BetaManagedAgentsDocumentBlock object`41914192 Document content, either specified directly as base64 data, as text, or as a reference via a URL.41934194 - `BetaManagedAgentsRedactedBlock object`41954196 Placeholder for content withheld by Anthropic model policy.41974198 - `processed_at: string`41994200 A timestamp in RFC 3339 format42014202 format: date-time42034204 - `to_session_thread_id: string`42054206 Public `sthr_` ID of the thread the message was sent to.42074208 - `to_agent_name: optional string or null`42094210 Name of the callable agent this message was sent to. Absent when sent to the primary agent.42114212 - `BetaManagedAgentsAgentThreadContextCompactedEvent object`42134214 Indicates that context compaction (summarization) occurred during the session.42154216 - `type: "agent.thread_context_compacted"`42174218 - `id: string`42194220 Unique identifier for this event.42214222 - `processed_at: string`42234224 A timestamp in RFC 3339 format42254226 format: date-time42274228 - `BetaManagedAgentsSessionErrorEvent object`42294230 An error event indicating a problem occurred during session execution.42314232 - `type: "session.error"`42334234 - `id: string`42354236 Unique identifier for this event.42374238 - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`42394240 - `BetaManagedAgentsUnknownError object`42414242 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.42434244 - `type: "unknown_error"`42454246 - `message: string`42474248 Human-readable error description.42494250 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`42514252 What the client should do next in response to this error.42534254 - `BetaManagedAgentsRetryStatusRetrying object`42554256 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.42574258 - `type: "retrying"`42594260 - `BetaManagedAgentsRetryStatusExhausted object`42614262 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.42634264 - `type: "exhausted"`42654266 - `BetaManagedAgentsRetryStatusTerminal object`42674268 The session encountered a terminal error and will transition to `terminated` state.42694270 - `type: "terminal"`42714272 - `BetaManagedAgentsModelOverloadedError object`42734274 The model is currently overloaded. Emitted after automatic retries are exhausted.42754276 - `type: "model_overloaded_error"`42774278 - `message: string`42794280 Human-readable error description.42814282 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`42834284 What the client should do next in response to this error.42854286 - `BetaManagedAgentsRetryStatusRetrying object`42874288 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.42894290 - `BetaManagedAgentsRetryStatusExhausted object`42914292 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.42934294 - `BetaManagedAgentsRetryStatusTerminal object`42954296 The session encountered a terminal error and will transition to `terminated` state.42974298 - `BetaManagedAgentsModelRateLimitedError object`42994300 The model request was rate-limited.43014302 - `type: "model_rate_limited_error"`43034304 - `message: string`43054306 Human-readable error description.43074308 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`43094310 What the client should do next in response to this error.43114312 - `BetaManagedAgentsRetryStatusRetrying object`43134314 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.43154316 - `BetaManagedAgentsRetryStatusExhausted object`43174318 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.43194320 - `BetaManagedAgentsRetryStatusTerminal object`43214322 The session encountered a terminal error and will transition to `terminated` state.43234324 - `BetaManagedAgentsModelRequestFailedError object`43254326 A model request failed for a reason other than overload or rate-limiting.43274328 - `type: "model_request_failed_error"`43294330 - `message: string`43314332 Human-readable error description.43334334 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`43354336 What the client should do next in response to this error.43374338 - `BetaManagedAgentsRetryStatusRetrying object`43394340 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.43414342 - `BetaManagedAgentsRetryStatusExhausted object`43434344 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.43454346 - `BetaManagedAgentsRetryStatusTerminal object`43474348 The session encountered a terminal error and will transition to `terminated` state.43494350 - `BetaManagedAgentsMCPConnectionFailedError object`43514352 Failed to connect to an MCP server.43534354 - `type: "mcp_connection_failed_error"`43554356 - `mcp_server_name: string`43574358 Name of the MCP server that failed to connect.43594360 - `message: string`43614362 Human-readable error description.43634364 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`43654366 What the client should do next in response to this error.43674368 - `BetaManagedAgentsRetryStatusRetrying object`43694370 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.43714372 - `BetaManagedAgentsRetryStatusExhausted object`43734374 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.43754376 - `BetaManagedAgentsRetryStatusTerminal object`43774378 The session encountered a terminal error and will transition to `terminated` state.43794380 - `BetaManagedAgentsMCPAuthenticationFailedError object`43814382 Authentication to an MCP server failed.43834384 - `type: "mcp_authentication_failed_error"`43854386 - `mcp_server_name: string`43874388 Name of the MCP server that failed authentication.43894390 - `message: string`43914392 Human-readable error description.43934394 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`43954396 What the client should do next in response to this error.43974398 - `BetaManagedAgentsRetryStatusRetrying object`43994400 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.44014402 - `BetaManagedAgentsRetryStatusExhausted object`44034404 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.44054406 - `BetaManagedAgentsRetryStatusTerminal object`44074408 The session encountered a terminal error and will transition to `terminated` state.44094410 - `BetaManagedAgentsBillingError object`44114412 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.44134414 - `type: "billing_error"`44154416 - `message: string`44174418 Human-readable error description.44194420 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`44214422 What the client should do next in response to this error.44234424 - `BetaManagedAgentsRetryStatusRetrying object`44254426 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.44274428 - `BetaManagedAgentsRetryStatusExhausted object`44294430 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.44314432 - `BetaManagedAgentsRetryStatusTerminal object`44334434 The session encountered a terminal error and will transition to `terminated` state.44354436 - `BetaManagedAgentsCredentialHostUnreachableError object`44374438 An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.44394440 - `type: "credential_host_unreachable_error"`44414442 - `credential_id: string`44434444 ID of the affected credential.44454446 - `message: string`44474448 Human-readable error description.44494450 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`44514452 What the client should do next in response to this error.44534454 - `BetaManagedAgentsRetryStatusRetrying object`44554456 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.44574458 - `BetaManagedAgentsRetryStatusExhausted object`44594460 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.44614462 - `BetaManagedAgentsRetryStatusTerminal object`44634464 The session encountered a terminal error and will transition to `terminated` state.44654466 - `vault_id: string`44674468 ID of the vault containing the affected credential.44694470 - `processed_at: string`44714472 A timestamp in RFC 3339 format44734474 format: date-time44754476 - `BetaManagedAgentsSessionStatusRescheduledEvent object`44774478 Indicates the session is recovering from an error state and is rescheduled for execution.44794480 - `type: "session.status_rescheduled"`44814482 - `id: string`44834484 Unique identifier for this event.44854486 - `processed_at: string`44874488 A timestamp in RFC 3339 format44894490 format: date-time44914492 - `BetaManagedAgentsSessionStatusRunningEvent object`44934494 Indicates the session is actively running and the agent is working.44954496 - `type: "session.status_running"`44974498 - `id: string`44994500 Unique identifier for this event.45014502 - `processed_at: string`45034504 A timestamp in RFC 3339 format45054506 format: date-time45074508 - `BetaManagedAgentsSessionStatusIdleEvent object`45094510 Indicates the agent has paused and is awaiting user input.45114512 - `type: "session.status_idle"`45134514 - `id: string`45154516 Unique identifier for this event.45174518 - `processed_at: string`45194520 A timestamp in RFC 3339 format45214522 format: date-time45234524 - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`45254526 - `BetaManagedAgentsSessionEndTurn object`45274528 The agent completed its turn naturally and is ready for the next user message.45294530 - `type: "end_turn"`45314532 - `BetaManagedAgentsSessionRequiresAction object`45334534 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.45354536 - `type: "requires_action"`45374538 - `event_ids: array of string`45394540 The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.45414542 - `BetaManagedAgentsSessionRetriesExhausted object`45434544 The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.45454546 - `type: "retries_exhausted"`45474548 - `BetaManagedAgentsSessionBudgetReached object`45494550 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.45514552 - `type: "budget_reached"`45534554 - `BetaManagedAgentsSessionStatusTerminatedEvent object`45554556 Indicates the session has terminated, either due to an error or completion.45574558 - `type: "session.status_terminated"`45594560 - `id: string`45614562 Unique identifier for this event.45634564 - `processed_at: string`45654566 A timestamp in RFC 3339 format45674568 format: date-time45694570 - `BetaManagedAgentsSessionThreadCreatedEvent object`45714572 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.45734574 - `type: "session.thread_created"`45754576 - `id: string`45774578 Unique identifier for this event.45794580 - `agent_name: string`45814582 Name of the callable agent the thread runs.45834584 - `processed_at: string`45854586 A timestamp in RFC 3339 format45874588 format: date-time45894590 - `session_thread_id: string`45914592 Public `sthr_` ID of the newly created thread.45934594 - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`45954596 Emitted when an outcome evaluation cycle begins.45974598 - `type: "span.outcome_evaluation_start"`45994600 - `id: string`46014602 Unique identifier for this event.46034604 - `iteration: number`46054606 0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.46074608 format: int3246094610 - `outcome_id: string`46114612 The `outc_` ID of the outcome being evaluated.46134614 - `processed_at: string`46154616 A timestamp in RFC 3339 format46174618 format: date-time46194620 - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`46214622 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.46234624 - `type: "span.outcome_evaluation_end"`46254626 - `id: string`46274628 Unique identifier for this event.46294630 - `explanation: string`46314632 Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.46334634 - `iteration: number`46354636 0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.46374638 format: int3246394640 - `outcome_evaluation_start_id: string`46414642 The id of the corresponding `span.outcome_evaluation_start` event.46434644 - `outcome_id: string`46454646 The `outc_` ID of the outcome being evaluated.46474648 - `processed_at: string`46494650 A timestamp in RFC 3339 format46514652 format: date-time46534654 - `result: string`46554656 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.46574658 - `usage: BetaManagedAgentsSpanModelUsage`46594660 Token usage for a single model request.46614662 - `cache_creation_input_tokens: number`46634664 Tokens used to create prompt cache in this request.46654666 format: int3246674668 - `cache_read_input_tokens: number`46694670 Tokens read from prompt cache in this request.46714672 format: int3246734674 - `input_tokens: number`46754676 Input tokens consumed by this request.46774678 format: int3246794680 - `output_tokens: number`46814682 Output tokens generated by this request.46834684 format: int3246854686 - `speed: optional "standard" or "fast" or null`46874688 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.46894690 - `"standard"`46914692 - `"fast"`46934694 - `BetaManagedAgentsSpanModelRequestStartEvent object`46954696 Emitted when a model request is initiated by the agent.46974698 - `type: "span.model_request_start"`46994700 - `id: string`47014702 Unique identifier for this event.47034704 - `processed_at: string`47054706 A timestamp in RFC 3339 format47074708 format: date-time47094710 - `BetaManagedAgentsSpanModelRequestEndEvent object`47114712 Emitted when a model request completes.47134714 - `type: "span.model_request_end"`47154716 - `id: string`47174718 Unique identifier for this event.47194720 - `is_error: boolean or null`47214722 Whether the model request resulted in an error.47234724 - `model_request_start_id: string`47254726 The id of the corresponding `span.model_request_start` event.47274728 - `model_usage: BetaManagedAgentsSpanModelUsage`47294730 Token usage for a single model request.47314732 - `processed_at: string`47334734 A timestamp in RFC 3339 format47354736 format: date-time47374738 - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`47394740 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.47414742 - `type: "span.outcome_evaluation_ongoing"`47434744 - `id: string`47454746 Unique identifier for this event.47474748 - `iteration: number`47494750 0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.47514752 format: int3247534754 - `outcome_id: string`47554756 The `outc_` ID of the outcome being evaluated.47574758 - `processed_at: string`47594760 A timestamp in RFC 3339 format47614762 format: date-time47634764 - `BetaManagedAgentsUserDefineOutcomeEvent object`47654766 Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.47674768 - `type: "user.define_outcome"`47694770 - `id: string`47714772 Unique identifier for this event.47734774 - `description: string`47754776 What the agent should produce. Copied from the input event.47774778 - `max_iterations: number or null`47794780 Evaluate-then-revise cycles before giving up. Default 3, max 20.47814782 format: int3247834784 - `outcome_id: string`47854786 Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.47874788 - `processed_at: string`47894790 A timestamp in RFC 3339 format47914792 format: date-time47934794 - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`47954796 Rubric for grading the quality of an outcome.47974798 - `BetaManagedAgentsFileRubric object`47994800 Rubric referenced by a file uploaded via the Files API.48014802 - `type: "file"`48034804 - `file_id: string`48054806 ID of the rubric file.48074808 - `BetaManagedAgentsTextRubric object`48094810 Rubric content provided inline as text.48114812 - `type: "text"`48134814 - `content: string`48154816 Rubric content. Plain text or markdown — the grader treats it as freeform text.48174818 - `BetaManagedAgentsSessionDeletedEvent object`48194820 Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.48214822 - `type: "session.deleted"`48234824 - `id: string`48254826 Unique identifier for this event.48274828 - `processed_at: string`48294830 A timestamp in RFC 3339 format48314832 format: date-time48334834 - `BetaManagedAgentsSessionThreadStatusRunningEvent object`48354836 A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.48374838 - `type: "session.thread_status_running"`48394840 - `id: string`48414842 Unique identifier for this event.48434844 - `agent_name: string`48454846 Name of the agent the thread runs.48474848 - `processed_at: string`48494850 A timestamp in RFC 3339 format48514852 format: date-time48534854 - `session_thread_id: string`48554856 Public sthr_ ID of the thread that started running.48574858 - `BetaManagedAgentsSessionThreadStatusIdleEvent object`48594860 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.48614862 - `type: "session.thread_status_idle"`48634864 - `id: string`48654866 Unique identifier for this event.48674868 - `agent_name: string`48694870 Name of the agent the thread runs.48714872 - `processed_at: string`48734874 A timestamp in RFC 3339 format48754876 format: date-time48774878 - `session_thread_id: string`48794880 Public sthr_ ID of the thread that went idle.48814882 - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`48834884 - `BetaManagedAgentsSessionEndTurn object`48854886 The agent completed its turn naturally and is ready for the next user message.48874888 - `BetaManagedAgentsSessionRequiresAction object`48894890 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.48914892 - `BetaManagedAgentsSessionRetriesExhausted object`48934894 The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.48954896 - `BetaManagedAgentsSessionBudgetReached object`48974898 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.48994900 - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`49014902 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.49034904 - `type: "session.thread_status_terminated"`49054906 - `id: string`49074908 Unique identifier for this event.49094910 - `agent_name: string`49114912 Name of the agent the thread runs.49134914 - `processed_at: string`49154916 A timestamp in RFC 3339 format49174918 format: date-time49194920 - `session_thread_id: string`49214922 Public sthr_ ID of the thread that terminated.49234924 - `BetaManagedAgentsUserToolResultEvent object`49254926 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.49274928 - `type: "user.tool_result"`49294930 - `id: string`49314932 Unique identifier for this event.49334934 - `tool_use_id: string`49354936 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.49374938 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`49394940 The result content returned by the tool.49414942 - `BetaManagedAgentsTextBlock object`49434944 Regular text content.49454946 - `BetaManagedAgentsImageBlock object`49474948 Image content specified directly as base64 data or as a reference via a URL.49494950 - `BetaManagedAgentsDocumentBlock object`49514952 Document content, either specified directly as base64 data, as text, or as a reference via a URL.49534954 - `BetaManagedAgentsSearchResultBlock object`49554956 A block containing a web search result.49574958 - `is_error: optional boolean or null`49594960 Whether the tool execution resulted in an error.49614962 - `processed_at: optional string or null`49634964 A timestamp in RFC 3339 format49654966 format: date-time49674968 - `session_thread_id: optional string or null`49694970 Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.49714972 - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`49734974 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.49754976 - `type: "session.thread_status_rescheduled"`49774978 - `id: string`49794980 Unique identifier for this event.49814982 - `agent_name: string`49834984 Name of the agent the thread runs.49854986 - `processed_at: string`49874988 A timestamp in RFC 3339 format49894990 format: date-time49914992 - `session_thread_id: string`49934994 Public sthr_ ID of the thread that is retrying.49954996 - `BetaManagedAgentsSessionUpdatedEvent object`49974998 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.49995000 - `type: "session.updated"`50015002 - `id: string`50035004 Unique identifier for this event.50055006 - `processed_at: string`50075008 A timestamp in RFC 3339 format50095010 format: date-time50115012 - `agent: optional BetaManagedAgentsSessionAgent or null`50135014 Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.50155016 - `type: "agent"`50175018 - `id: string`50195020 - `description: string or null`50215022 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`50235024 - `type: "url"`50255026 - `name: string`50275028 - `url: string`50295030 - `model: BetaManagedAgentsModelConfig`50315032 Model identifier and configuration.50335034 - `id: BetaManagedAgentsModel`50355036 The model that will power your agent.50375038 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.50395040 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`50415042 The model that will power your agent.50435044 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.50455046 - `"claude-fable-5-1"`50475048 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows50495050 - `"claude-sonnet-5"`50515052 High-performance model for coding and agents50535054 - `"claude-fable-5"`50555056 Next generation of intelligence for the hardest knowledge work and coding problems50575058 - `"claude-opus-5"`50595060 Powerful intelligence for long-running agents and coding50615062 - `"claude-opus-4-8"`50635064 Powerful intelligence for long-running agents and coding50655066 - `"claude-opus-4-7"`50675068 Powerful intelligence for long-running agents and coding50695070 - `"claude-opus-4-6"`50715072 Powerful intelligence for long-running agents and coding50735074 - `"claude-sonnet-4-6"`50755076 Best combination of speed and intelligence50775078 - `"claude-haiku-4-5"`50795080 Fastest model with near-frontier intelligence50815082 - `"claude-haiku-4-5-20251001"`50835084 Fastest model with near-frontier intelligence50855086 - `"claude-opus-4-5"`50875088 Powerful intelligence for long-running agents and coding50895090 - `"claude-opus-4-5-20251101"`50915092 Powerful intelligence for long-running agents and coding50935094 - `"claude-sonnet-4-5"`50955096 High-performance model for agents and coding50975098 - `"claude-sonnet-4-5-20250929"`50995100 High-performance model for agents and coding51015102 - `string`51035104 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`51055106 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.51075108 - `BetaManagedAgentsEffortLow object`51095110 Low effort. Favors latency over reasoning depth.51115112 - `type: "low"`51135114 - `BetaManagedAgentsEffortMedium object`51155116 Medium effort. Balances latency and reasoning depth.51175118 - `type: "medium"`51195120 - `BetaManagedAgentsEffortHigh object`51215122 High effort. Favors reasoning depth.51235124 - `type: "high"`51255126 - `BetaManagedAgentsEffortXhigh object`51275128 Extra-high effort. Not all models accept this level.51295130 - `type: "xhigh"`51315132 - `BetaManagedAgentsEffortMax object`51335134 Maximum effort. Favors reasoning depth over latency.51355136 - `type: "max"`51375138 - `inference_geo: optional string`51395140 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.51415142 - `speed: optional "standard" or "fast"`51435144 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.51455146 - `"standard"`51475148 - `"fast"`51495150 - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`51515152 Resolved coordinator topology with full agent definitions for each roster member.51535154 - `type: "coordinator"`51555156 - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`51575158 Full `agent` definitions the coordinator may spawn as session threads.51595160 - `BetaManagedAgentsSessionThreadAgent object`51615162 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`.51635164 - `type: "agent"`51655166 - `id: string`51675168 - `description: string or null`51695170 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`51715172 - `type: "url"`51735174 - `name: string`51755176 - `url: string`51775178 - `model: BetaManagedAgentsModelConfig`51795180 Model identifier and configuration.51815182 - `name: string`51835184 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`51855186 - `BetaManagedAgentsAnthropicSkill object`51875188 A resolved Anthropic-managed skill.51895190 - `type: "anthropic"`51915192 - `skill_id: string`51935194 - `version: string`51955196 - `BetaManagedAgentsCustomSkill object`51975198 A resolved user-created custom skill.51995200 - `type: "custom"`52015202 - `skill_id: string`52035204 - `version: string`52055206 - `system: string or null`52075208 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`52095210 - `BetaManagedAgentsAgentToolset20260401 object`52115212 - `type: "agent_toolset_20260401"`52135214 - `configs: array of BetaManagedAgentsAgentToolConfig`52155216 - `BetaManagedAgentsBashToolConfig object`52175218 Configuration for the bash tool.52195220 - `type: "bash"`52215222 - `enabled: boolean`52235224 - `name: "bash"`52255226 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`52275228 Permission policy for tool execution.52295230 - `BetaManagedAgentsAlwaysAllowPolicy object`52315232 Tool calls are automatically approved without user confirmation.52335234 - `type: "always_allow"`52355236 - `BetaManagedAgentsAlwaysAskPolicy object`52375238 Tool calls require user confirmation before execution.52395240 - `type: "always_ask"`52415242 - `BetaManagedAgentsAutoPolicy object`52435244 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.52455246 - `type: "auto"`52475248 - `BetaManagedAgentsEditToolConfig object`52495250 Configuration for the edit tool.52515252 - `type: "edit"`52535254 - `enabled: boolean`52555256 - `name: "edit"`52575258 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`52595260 Permission policy for tool execution.52615262 - `BetaManagedAgentsAlwaysAllowPolicy object`52635264 Tool calls are automatically approved without user confirmation.52655266 - `BetaManagedAgentsAlwaysAskPolicy object`52675268 Tool calls require user confirmation before execution.52695270 - `BetaManagedAgentsAutoPolicy object`52715272 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.52735274 - `BetaManagedAgentsReadToolConfig object`52755276 Configuration for the read tool.52775278 - `type: "read"`52795280 - `enabled: boolean`52815282 - `name: "read"`52835284 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`52855286 Permission policy for tool execution.52875288 - `BetaManagedAgentsAlwaysAllowPolicy object`52895290 Tool calls are automatically approved without user confirmation.52915292 - `BetaManagedAgentsAlwaysAskPolicy object`52935294 Tool calls require user confirmation before execution.52955296 - `BetaManagedAgentsAutoPolicy object`52975298 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.52995300 - `BetaManagedAgentsWriteToolConfig object`53015302 Configuration for the write tool.53035304 - `type: "write"`53055306 - `enabled: boolean`53075308 - `name: "write"`53095310 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`53115312 Permission policy for tool execution.53135314 - `BetaManagedAgentsAlwaysAllowPolicy object`53155316 Tool calls are automatically approved without user confirmation.53175318 - `BetaManagedAgentsAlwaysAskPolicy object`53195320 Tool calls require user confirmation before execution.53215322 - `BetaManagedAgentsAutoPolicy object`53235324 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.53255326 - `BetaManagedAgentsGlobToolConfig object`53275328 Configuration for the glob tool.53295330 - `type: "glob"`53315332 - `enabled: boolean`53335334 - `name: "glob"`53355336 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`53375338 Permission policy for tool execution.53395340 - `BetaManagedAgentsAlwaysAllowPolicy object`53415342 Tool calls are automatically approved without user confirmation.53435344 - `BetaManagedAgentsAlwaysAskPolicy object`53455346 Tool calls require user confirmation before execution.53475348 - `BetaManagedAgentsAutoPolicy object`53495350 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.53515352 - `BetaManagedAgentsGrepToolConfig object`53535354 Configuration for the grep tool.53555356 - `type: "grep"`53575358 - `enabled: boolean`53595360 - `name: "grep"`53615362 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`53635364 Permission policy for tool execution.53655366 - `BetaManagedAgentsAlwaysAllowPolicy object`53675368 Tool calls are automatically approved without user confirmation.53695370 - `BetaManagedAgentsAlwaysAskPolicy object`53715372 Tool calls require user confirmation before execution.53735374 - `BetaManagedAgentsAutoPolicy object`53755376 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.53775378 - `BetaManagedAgentsWebFetchToolConfig object`53795380 Configuration for the web_fetch tool.53815382 - `type: "web_fetch"`53835384 - `enabled: boolean`53855386 - `name: "web_fetch"`53875388 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`53895390 Permission policy for tool execution.53915392 - `BetaManagedAgentsAlwaysAllowPolicy object`53935394 Tool calls are automatically approved without user confirmation.53955396 - `BetaManagedAgentsAlwaysAskPolicy object`53975398 Tool calls require user confirmation before execution.53995400 - `BetaManagedAgentsAutoPolicy object`54015402 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.54035404 - `allowed_domains: optional array of string`54055406 - `blocked_domains: optional array of string`54075408 - `max_content_tokens: optional number or null`54095410 format: int3254115412 - `BetaManagedAgentsWebSearchToolConfig object`54135414 Configuration for the web_search tool.54155416 - `type: "web_search"`54175418 - `enabled: boolean`54195420 - `name: "web_search"`54215422 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`54235424 Permission policy for tool execution.54255426 - `BetaManagedAgentsAlwaysAllowPolicy object`54275428 Tool calls are automatically approved without user confirmation.54295430 - `BetaManagedAgentsAlwaysAskPolicy object`54315432 Tool calls require user confirmation before execution.54335434 - `BetaManagedAgentsAutoPolicy object`54355436 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.54375438 - `allowed_domains: optional array of string`54395440 - `blocked_domains: optional array of string`54415442 - `user_location: optional BetaManagedAgentsUserLocation or null`54435444 Approximate user location for search result localization.54455446 - `type: "approximate"`54475448 Location precision. Only "approximate" is supported.54495450 - `city: optional string or null`54515452 City name.54535454 minLength: 1, maxLength: 25554555456 - `country: optional string or null`54575458 Two-letter ISO 3166-1 country code, uppercase.54595460 - `region: optional string or null`54615462 Region or state name.54635464 minLength: 1, maxLength: 25554655466 - `timezone: optional string or null`54675468 IANA timezone identifier, e.g. "America/Los_Angeles".54695470 minLength: 1, maxLength: 25554715472 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`54735474 Resolved default configuration for agent tools.54755476 - `enabled: boolean`54775478 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`54795480 Permission policy for tool execution.54815482 - `BetaManagedAgentsAlwaysAllowPolicy object`54835484 Tool calls are automatically approved without user confirmation.54855486 - `BetaManagedAgentsAlwaysAskPolicy object`54875488 Tool calls require user confirmation before execution.54895490 - `BetaManagedAgentsAutoPolicy object`54915492 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.54935494 - `BetaManagedAgentsMCPToolset object`54955496 - `type: "mcp_toolset"`54975498 - `configs: array of BetaManagedAgentsMCPToolConfig`54995500 - `enabled: boolean`55015502 - `name: string`55035504 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`55055506 Permission policy for tool execution.55075508 - `BetaManagedAgentsAlwaysAllowPolicy object`55095510 Tool calls are automatically approved without user confirmation.55115512 - `BetaManagedAgentsAlwaysAskPolicy object`55135514 Tool calls require user confirmation before execution.55155516 - `BetaManagedAgentsAutoPolicy object`55175518 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.55195520 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`55215522 Resolved default configuration for all tools from an MCP server.55235524 - `enabled: boolean`55255526 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`55275528 Permission policy for tool execution.55295530 - `BetaManagedAgentsAlwaysAllowPolicy object`55315532 Tool calls are automatically approved without user confirmation.55335534 - `BetaManagedAgentsAlwaysAskPolicy object`55355536 Tool calls require user confirmation before execution.55375538 - `BetaManagedAgentsAutoPolicy object`55395540 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.55415542 - `mcp_server_name: string`55435544 - `BetaManagedAgentsCustomTool object`55455546 A custom tool as returned in API responses.55475548 - `type: "custom"`55495550 - `description: string`55515552 - `input_schema: BetaManagedAgentsCustomToolInputSchema`55535554 JSON Schema for custom tool input parameters.55555556 - `type: "object"`55575558 - `properties: optional map[unknown] or null`55595560 - `required: optional array of string or null`55615562 - `name: string`55635564 - `version: number`55655566 format: int3255675568 - `BetaManagedAgentsAdvisor object`55695570 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.55715572 - `type: "advisor"`55735574 - `model: string`55755576 The advisor model id.55775578 - `name: string`55795580 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`55815582 - `BetaManagedAgentsAnthropicSkill object`55835584 A resolved Anthropic-managed skill.55855586 - `BetaManagedAgentsCustomSkill object`55875588 A resolved user-created custom skill.55895590 - `system: string or null`55915592 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`55935594 - `BetaManagedAgentsAgentToolset20260401 object`55955596 - `BetaManagedAgentsMCPToolset object`55975598 - `BetaManagedAgentsCustomTool object`55995600 A custom tool as returned in API responses.56015602 - `version: number`56035604 format: int3256055606 - `budget: optional BetaManagedAgentsBudgetLimit or null`56075608 A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.56095610 - `type: "limit"`56115612 - `max_list_cost: BetaMonetaryAmount`56135614 A monetary amount in a specific currency.56155616 - `amount: string`56175618 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.56195620 - `currency: BetaCurrency`56215622 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.56235624 - `metadata: optional map[string]`56255626 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.56275628 - `title: optional string or null`56295630 The session's new title. Present only when the update changed it.56315632 - `BetaManagedAgentsStartEvent object`56335634 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.56355636 - `type: "event_start"`56375638 - `event: BetaManagedAgentsStartEventPreview`56395640 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.56415642 - `BetaManagedAgentsAgentMessagePreview object`56435644 - `type: "agent.message"`56455646 - `id: string`56475648 The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.56495650 - `BetaManagedAgentsAgentThinkingPreview object`56515652 - `type: "agent.thinking"`56535654 - `id: string`56555656 The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.56575658 - `BetaManagedAgentsDeltaEvent object`56595660 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.56615662 - `type: "event_delta"`56635664 - `delta: BetaManagedAgentsDeltaContent`56655666 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.56675668 - `type: "content_delta"`56695670 - `content: BetaManagedAgentsTextBlock`56715672 Regular text content.56735674 - `index: optional number`56755676 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.56775678 format: uint3256795680 - `event_id: string`56815682 The id of the event being previewed. Matches event.id on the corresponding event_start and the buffered event that reconciles the preview.56835684 - `BetaManagedAgentsSystemMessageEvent object`56855686 A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.56875688 - `type: "system.message"`56895690 - `id: string`56915692 Unique identifier for this event.56935694 - `content: array of BetaManagedAgentsSystemContentBlock`56955696 System content blocks. Text-only.56975698 - `type: "text"`56995700 - `text: string`57015702 The text content.57035704 minLength: 157055706 - `processed_at: optional string or null`57075708 A timestamp in RFC 3339 format57095710 format: date-time57115712 - `BetaManagedAgentsSessionUsageEvent object`57135714 Periodic snapshot of the session's cumulative usage and tracked list cost.57155716 - `type: "session.usage"`57175718 - `id: string`57195720 Unique identifier for this event.57215722 - `processed_at: string`57235724 A timestamp in RFC 3339 format57255726 format: date-time57275728 - `usage: BetaManagedAgentsSessionUsageSnapshot`57295730 Point-in-time snapshot of a session's cumulative usage.57315732 - `active_seconds: optional number`57335734 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.57355736 format: double57375738 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`57395740 Prompt-cache creation token usage broken down by cache lifetime.57415742 - `ephemeral_1h_input_tokens: optional number`57435744 Tokens used to create 1-hour ephemeral cache entries.57455746 format: int3257475748 - `ephemeral_5m_input_tokens: optional number`57495750 Tokens used to create 5-minute ephemeral cache entries.57515752 format: int3257535754 - `cache_read_input_tokens: optional number`57555756 Total tokens read from prompt cache.57575758 format: int3257595760 - `input_tokens: optional number`57615762 Total input tokens consumed across all turns.57635764 format: int3257655766 - `list_cost: optional BetaMonetaryAmount`57675768 A monetary amount in a specific currency.57695770 - `output_tokens: optional number`57715772 Total output tokens generated across all turns.57735774 format: int3257755776 - `server_tool_use: optional BetaManagedAgentsServerToolUsage`57775778 Cumulative count of server-executed tool invocations, broken down by tool.57795780 - `web_fetch_requests: optional number`57815782 Number of server-executed web fetch requests.57835784 format: int3257855786 - `web_search_requests: optional number`57875788 Number of server-executed web search requests.57895790 format: int3257915792 - `budget: optional BetaManagedAgentsBudgetLimit or null`57935794 A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.57955796## Threads › Events57975798### List Session Thread Events57995800**GET** `/v1/sessions/{session_id}/threads/{thread_id}/events`58015802List Session Thread Events58035804#### Path parameters58055806- `session_id: string`58075808- `thread_id: string`58095810#### Query parameters58115812- `limit: optional number`58135814 Query parameter for limit58155816 format: int3258175818- `page: optional string`58195820 Query parameter for page58215822#### Headers58235824- `"anthropic-beta": optional array of AnthropicBeta`58255826 Optional header to specify the beta version(s) you want to use.58275828 - `string`58295830 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`58315832 - `"message-batches-2024-09-24"`58335834 - `"prompt-caching-2024-07-31"`58355836 - `"computer-use-2024-10-22"`58375838 - `"computer-use-2025-01-24"`58395840 - `"pdfs-2024-09-25"`58415842 - `"token-counting-2024-11-01"`58435844 - `"token-efficient-tools-2025-02-19"`58455846 - `"output-128k-2025-02-19"`58475848 - `"files-api-2025-04-14"`58495850 - `"mcp-client-2025-04-04"`58515852 - `"mcp-client-2025-11-20"`58535854 - `"dev-full-thinking-2025-05-14"`58555856 - `"interleaved-thinking-2025-05-14"`58575858 - `"code-execution-2025-05-22"`58595860 - `"extended-cache-ttl-2025-04-11"`58615862 - `"context-1m-2025-08-07"`58635864 - `"context-management-2025-06-27"`58655866 - `"model-context-window-exceeded-2025-08-26"`58675868 - `"skills-2025-10-02"`58695870 - `"fast-mode-2026-02-01"`58715872 - `"output-300k-2026-03-24"`58735874 - `"user-profiles-2026-03-24"`58755876 - `"user-profiles-2026-08-18"`58775878 - `"user-profiles-2026-09-04"`58795880 - `"advisor-tool-2026-03-01"`58815882 - `"managed-agents-2026-04-01"`58835884 - `"cache-diagnosis-2026-04-07"`58855886 - `"dreaming-2026-04-21"`58875888 - `"thinking-token-count-2026-05-13"`58895890 - `"server-side-fallback-2026-06-01"`58915892 - `"server-side-fallback-2026-07-01"`58935894 - `"fallback-credit-2026-06-01"`58955896 - `"fallback-credit-2026-07-01"`58975898 - `"agent-memory-2026-07-22"`58995900 - `"mid-conversation-tool-changes-2026-07-01"`59015902 - `"compact-2026-01-12"`59035904 - `"computer-use-2025-11-24"`59055906 - `"mcp-tunnels-2026-06-22"`59075908 - `"structured-outputs-2025-11-13"`59095910 - `"task-budgets-2026-03-13"`59115912 - `"thinking-display-updates-2026-08-18"`59135914 - `"ce-user-management-2026-07-13"`59155916 - `"mid-conversation-output-config-2026-07-01"`59175918 - `"thinking-binding-controls-2026-08-01"`59195920 - `"mid-conversation-system-clear-at-2026-08-21"`59215922 - `"compact-2026-09-04"`59235924- `"anthropic-workspace-id": optional string`59255926#### Returns59275928- `data: optional array of BetaManagedAgentsSessionEvent`59295930 Events for the thread, ordered by `processed_at`.59315932 - `BetaManagedAgentsUserMessageEvent object`59335934 A user message event in the session conversation.59355936 - `type: "user.message"`59375938 - `id: string`59395940 Unique identifier for this event.59415942 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`59435944 Array of content blocks comprising the user message.59455946 - `BetaManagedAgentsTextBlock object`59475948 Regular text content.59495950 - `type: "text"`59515952 - `text: string`59535954 The text content.59555956 minLength: 159575958 - `BetaManagedAgentsImageBlock object`59595960 Image content specified directly as base64 data or as a reference via a URL.59615962 - `type: "image"`59635964 - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`59655966 Union type for image source variants.59675968 - `BetaManagedAgentsBase64ImageSource object`59695970 Base64-encoded image data.59715972 - `type: "base64"`59735974 - `data: string`59755976 Base64-encoded image data.59775978 minLength: 159795980 - `media_type: string`59815982 MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").59835984 minLength: 159855986 - `BetaManagedAgentsURLImageSource object`59875988 Image referenced by URL.59895990 - `type: "url"`59915992 - `url: string`59935994 URL of the image to fetch.59955996 minLength: 159975998 - `BetaManagedAgentsFileImageSource object`59996000 Image referenced by file ID.60016002 - `type: "file"`60036004 - `file_id: string`60056006 ID of a previously uploaded file.60076008 minLength: 160096010 - `BetaManagedAgentsDocumentBlock object`60116012 Document content, either specified directly as base64 data, as text, or as a reference via a URL.60136014 - `type: "document"`60156016 - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`60176018 Union type for document source variants.60196020 - `BetaManagedAgentsBase64DocumentSource object`60216022 Base64-encoded document data.60236024 - `type: "base64"`60256026 - `data: string`60276028 Base64-encoded document data.60296030 minLength: 160316032 - `media_type: string`60336034 MIME type of the document (e.g., "application/pdf").60356036 minLength: 160376038 - `BetaManagedAgentsPlainTextDocumentSource object`60396040 Plain text document content.60416042 - `type: "text"`60436044 - `data: string`60456046 The plain text content.60476048 minLength: 160496050 - `media_type: "text/plain"`60516052 MIME type of the text content. Must be "text/plain".60536054 - `BetaManagedAgentsURLDocumentSource object`60556056 Document referenced by URL.60576058 - `type: "url"`60596060 - `url: string`60616062 URL of the document to fetch.60636064 minLength: 160656066 - `BetaManagedAgentsFileDocumentSource object`60676068 Document referenced by file ID.60696070 - `type: "file"`60716072 - `file_id: string`60736074 ID of a previously uploaded file.60756076 minLength: 160776078 - `context: optional string or null`60796080 Additional context about the document for the model.60816082 - `title: optional string or null`60836084 The title of the document.60856086 - `BetaManagedAgentsRedactedBlock object`60876088 Placeholder for content withheld by Anthropic model policy.60896090 - `type: "redacted"`60916092 - `processed_at: optional string or null`60936094 A timestamp in RFC 3339 format60956096 format: date-time60976098 - `BetaManagedAgentsUserInterruptEvent object`60996100 An interrupt event that pauses agent execution and returns control to the user.61016102 - `type: "user.interrupt"`61036104 - `id: string`61056106 Unique identifier for this event.61076108 - `processed_at: optional string or null`61096110 A timestamp in RFC 3339 format61116112 format: date-time61136114 - `session_thread_id: optional string or null`61156116 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.61176118 - `BetaManagedAgentsUserToolConfirmationEvent object`61196120 A tool confirmation event that approves or denies a pending tool execution.61216122 - `type: "user.tool_confirmation"`61236124 - `id: string`61256126 Unique identifier for this event.61276128 - `result: "allow" or "deny"`61296130 UserToolConfirmationResult enum61316132 - `"allow"`61336134 - `"deny"`61356136 - `tool_use_id: string`61376138 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.61396140 - `deny_message: optional string or null`61416142 Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.61436144 maxLength: 1000061456146 - `processed_at: optional string or null`61476148 A timestamp in RFC 3339 format61496150 format: date-time61516152 - `session_thread_id: optional string or null`61536154 Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.61556156 - `BetaManagedAgentsUserCustomToolResultEvent object`61576158 Event sent by the client providing the result of a custom tool execution.61596160 - `type: "user.custom_tool_result"`61616162 - `id: string`61636164 Unique identifier for this event.61656166 - `custom_tool_use_id: string`61676168 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.61696170 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`61716172 The result content returned by the tool.61736174 - `BetaManagedAgentsTextBlock object`61756176 Regular text content.61776178 - `BetaManagedAgentsImageBlock object`61796180 Image content specified directly as base64 data or as a reference via a URL.61816182 - `BetaManagedAgentsDocumentBlock object`61836184 Document content, either specified directly as base64 data, as text, or as a reference via a URL.61856186 - `BetaManagedAgentsSearchResultBlock object`61876188 A block containing a web search result.61896190 - `type: "search_result"`61916192 - `citations: BetaManagedAgentsSearchResultCitations`61936194 Citation settings for a search result.61956196 - `enabled: boolean`61976198 Whether citations are enabled for this search result.61996200 - `content: array of BetaManagedAgentsSearchResultContent`62016202 Array of text content blocks from the search result.62036204 - `type: "text"`62056206 - `text: string`62076208 The text content.62096210 minLength: 162116212 - `source: string`62136214 The URL source of the search result.62156216 minLength: 162176218 - `title: string`62196220 The title of the search result.62216222 minLength: 162236224 - `is_error: optional boolean or null`62256226 Whether the tool execution resulted in an error.62276228 - `processed_at: optional string or null`62296230 A timestamp in RFC 3339 format62316232 format: date-time62336234 - `session_thread_id: optional string or null`62356236 Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.62376238 - `BetaManagedAgentsAgentCustomToolUseEvent object`62396240 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.62416242 - `type: "agent.custom_tool_use"`62436244 - `id: string`62456246 Unique identifier for this event.62476248 - `input: map[unknown]`62496250 Input parameters for the tool call.62516252 - `name: string`62536254 Name of the custom tool being called.62556256 - `processed_at: string`62576258 A timestamp in RFC 3339 format62596260 format: date-time62616262 - `session_thread_id: optional string or null`62636264 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.62656266 - `BetaManagedAgentsAgentMessageEvent object`62676268 An agent response event in the session conversation.62696270 - `type: "agent.message"`62716272 - `id: string`62736274 Unique identifier for this event.62756276 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`62776278 Array of text blocks comprising the agent response.62796280 - `BetaManagedAgentsTextBlock object`62816282 Regular text content.62836284 - `BetaManagedAgentsRedactedBlock object`62856286 Placeholder for content withheld by Anthropic model policy.62876288 - `processed_at: string`62896290 A timestamp in RFC 3339 format62916292 format: date-time62936294 - `BetaManagedAgentsAgentThinkingEvent object`62956296 Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.62976298 - `type: "agent.thinking"`62996300 - `id: string`63016302 Unique identifier for this event.63036304 - `processed_at: string`63056306 A timestamp in RFC 3339 format63076308 format: date-time63096310 - `BetaManagedAgentsAgentMCPToolUseEvent object`63116312 Event emitted when the agent invokes a tool provided by an MCP server.63136314 - `type: "agent.mcp_tool_use"`63156316 - `id: string`63176318 Unique identifier for this event.63196320 - `input: map[unknown]`63216322 Input parameters for the tool call.63236324 - `mcp_server_name: string`63256326 Name of the MCP server providing the tool.63276328 - `name: string`63296330 Name of the MCP tool being used.63316332 - `processed_at: string`63336334 A timestamp in RFC 3339 format63356336 format: date-time63376338 - `evaluated_permission: optional "allow" or "ask" or "deny"`63396340 AgentEvaluatedPermission enum63416342 - `"allow"`63436344 - `"ask"`63456346 - `"deny"`63476348 - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`63496350 Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.63516352 - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`63536354 The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".63556356 - `type: "always_allow"`63576358 - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`63596360 The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".63616362 - `type: "always_ask"`63636364 - `BetaManagedAgentsAgentToolEvaluationAuto object`63656366 The resolved permission_policy was auto: the server judged this invocation individually.63676368 - `type: "auto"`63696370 - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`63716372 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.63736374 - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`63756376 The server judged the invocation safe to execute without client approval.63776378 - `type: "allow"`63796380 - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`63816382 The server reached no judgement; the invocation is held for client approval.63836384 - `type: "ask"`63856386 - `reason_code: string`63876388 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.63896390 maxLength: 6463916392 - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`63936394 The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.63956396 - `type: "deny"`63976398 - `reason_code: string`63996400 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.64016402 maxLength: 6464036404 - `session_thread_id: optional string or null`64056406 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.64076408 - `BetaManagedAgentsAgentMCPToolResultEvent object`64096410 Event representing the result of an MCP tool execution.64116412 - `type: "agent.mcp_tool_result"`64136414 - `id: string`64156416 Unique identifier for this event.64176418 - `mcp_tool_use_id: string`64196420 The id of the `agent.mcp_tool_use` event this result corresponds to.64216422 - `processed_at: string`64236424 A timestamp in RFC 3339 format64256426 format: date-time64276428 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`64296430 The result content returned by the tool.64316432 - `BetaManagedAgentsTextBlock object`64336434 Regular text content.64356436 - `BetaManagedAgentsImageBlock object`64376438 Image content specified directly as base64 data or as a reference via a URL.64396440 - `BetaManagedAgentsDocumentBlock object`64416442 Document content, either specified directly as base64 data, as text, or as a reference via a URL.64436444 - `BetaManagedAgentsSearchResultBlock object`64456446 A block containing a web search result.64476448 - `is_error: optional boolean or null`64496450 Whether the tool execution resulted in an error.64516452 - `BetaManagedAgentsAgentToolUseEvent object`64536454 Event emitted when the agent invokes a built-in agent tool.64556456 - `type: "agent.tool_use"`64576458 - `id: string`64596460 Unique identifier for this event.64616462 - `input: map[unknown]`64636464 Input parameters for the tool call.64656466 - `name: string`64676468 Name of the agent tool being used.64696470 - `processed_at: string`64716472 A timestamp in RFC 3339 format64736474 format: date-time64756476 - `evaluated_permission: optional "allow" or "ask" or "deny"`64776478 AgentEvaluatedPermission enum64796480 - `"allow"`64816482 - `"ask"`64836484 - `"deny"`64856486 - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`64876488 Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.64896490 - `session_thread_id: optional string or null`64916492 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.64936494 - `BetaManagedAgentsAgentToolResultEvent object`64956496 Event representing the result of an agent tool execution.64976498 - `type: "agent.tool_result"`64996500 - `id: string`65016502 Unique identifier for this event.65036504 - `processed_at: string`65056506 A timestamp in RFC 3339 format65076508 format: date-time65096510 - `tool_use_id: string`65116512 The id of the `agent.tool_use` event this result corresponds to.65136514 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`65156516 The result content returned by the tool.65176518 - `BetaManagedAgentsTextBlock object`65196520 Regular text content.65216522 - `BetaManagedAgentsImageBlock object`65236524 Image content specified directly as base64 data or as a reference via a URL.65256526 - `BetaManagedAgentsDocumentBlock object`65276528 Document content, either specified directly as base64 data, as text, or as a reference via a URL.65296530 - `BetaManagedAgentsSearchResultBlock object`65316532 A block containing a web search result.65336534 - `is_error: optional boolean or null`65356536 Whether the tool execution resulted in an error.65376538 - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`65396540 Delivery event written to the target thread's input stream when an agent-to-agent message arrives.65416542 - `type: "agent.thread_message_received"`65436544 - `id: string`65456546 Unique identifier for this event.65476548 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`65496550 Message content blocks.65516552 - `BetaManagedAgentsTextBlock object`65536554 Regular text content.65556556 - `BetaManagedAgentsImageBlock object`65576558 Image content specified directly as base64 data or as a reference via a URL.65596560 - `BetaManagedAgentsDocumentBlock object`65616562 Document content, either specified directly as base64 data, as text, or as a reference via a URL.65636564 - `BetaManagedAgentsRedactedBlock object`65656566 Placeholder for content withheld by Anthropic model policy.65676568 - `from_session_thread_id: string`65696570 Public `sthr_` ID of the thread that sent the message.65716572 - `processed_at: string`65736574 A timestamp in RFC 3339 format65756576 format: date-time65776578 - `from_agent_name: optional string or null`65796580 Name of the callable agent this message came from. Absent when received from the primary agent.65816582 - `BetaManagedAgentsAgentThreadMessageSentEvent object`65836584 Observability event emitted to the sender's output stream when an agent-to-agent message is sent.65856586 - `type: "agent.thread_message_sent"`65876588 - `id: string`65896590 Unique identifier for this event.65916592 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`65936594 Message content blocks.65956596 - `BetaManagedAgentsTextBlock object`65976598 Regular text content.65996600 - `BetaManagedAgentsImageBlock object`66016602 Image content specified directly as base64 data or as a reference via a URL.66036604 - `BetaManagedAgentsDocumentBlock object`66056606 Document content, either specified directly as base64 data, as text, or as a reference via a URL.66076608 - `BetaManagedAgentsRedactedBlock object`66096610 Placeholder for content withheld by Anthropic model policy.66116612 - `processed_at: string`66136614 A timestamp in RFC 3339 format66156616 format: date-time66176618 - `to_session_thread_id: string`66196620 Public `sthr_` ID of the thread the message was sent to.66216622 - `to_agent_name: optional string or null`66236624 Name of the callable agent this message was sent to. Absent when sent to the primary agent.66256626 - `BetaManagedAgentsAgentThreadContextCompactedEvent object`66276628 Indicates that context compaction (summarization) occurred during the session.66296630 - `type: "agent.thread_context_compacted"`66316632 - `id: string`66336634 Unique identifier for this event.66356636 - `processed_at: string`66376638 A timestamp in RFC 3339 format66396640 format: date-time66416642 - `BetaManagedAgentsSessionErrorEvent object`66436644 An error event indicating a problem occurred during session execution.66456646 - `type: "session.error"`66476648 - `id: string`66496650 Unique identifier for this event.66516652 - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`66536654 - `BetaManagedAgentsUnknownError object`66556656 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.66576658 - `type: "unknown_error"`66596660 - `message: string`66616662 Human-readable error description.66636664 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`66656666 What the client should do next in response to this error.66676668 - `BetaManagedAgentsRetryStatusRetrying object`66696670 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.66716672 - `type: "retrying"`66736674 - `BetaManagedAgentsRetryStatusExhausted object`66756676 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.66776678 - `type: "exhausted"`66796680 - `BetaManagedAgentsRetryStatusTerminal object`66816682 The session encountered a terminal error and will transition to `terminated` state.66836684 - `type: "terminal"`66856686 - `BetaManagedAgentsModelOverloadedError object`66876688 The model is currently overloaded. Emitted after automatic retries are exhausted.66896690 - `type: "model_overloaded_error"`66916692 - `message: string`66936694 Human-readable error description.66956696 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`66976698 What the client should do next in response to this error.66996700 - `BetaManagedAgentsRetryStatusRetrying object`67016702 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.67036704 - `BetaManagedAgentsRetryStatusExhausted object`67056706 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.67076708 - `BetaManagedAgentsRetryStatusTerminal object`67096710 The session encountered a terminal error and will transition to `terminated` state.67116712 - `BetaManagedAgentsModelRateLimitedError object`67136714 The model request was rate-limited.67156716 - `type: "model_rate_limited_error"`67176718 - `message: string`67196720 Human-readable error description.67216722 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`67236724 What the client should do next in response to this error.67256726 - `BetaManagedAgentsRetryStatusRetrying object`67276728 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.67296730 - `BetaManagedAgentsRetryStatusExhausted object`67316732 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.67336734 - `BetaManagedAgentsRetryStatusTerminal object`67356736 The session encountered a terminal error and will transition to `terminated` state.67376738 - `BetaManagedAgentsModelRequestFailedError object`67396740 A model request failed for a reason other than overload or rate-limiting.67416742 - `type: "model_request_failed_error"`67436744 - `message: string`67456746 Human-readable error description.67476748 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`67496750 What the client should do next in response to this error.67516752 - `BetaManagedAgentsRetryStatusRetrying object`67536754 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.67556756 - `BetaManagedAgentsRetryStatusExhausted object`67576758 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.67596760 - `BetaManagedAgentsRetryStatusTerminal object`67616762 The session encountered a terminal error and will transition to `terminated` state.67636764 - `BetaManagedAgentsMCPConnectionFailedError object`67656766 Failed to connect to an MCP server.67676768 - `type: "mcp_connection_failed_error"`67696770 - `mcp_server_name: string`67716772 Name of the MCP server that failed to connect.67736774 - `message: string`67756776 Human-readable error description.67776778 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`67796780 What the client should do next in response to this error.67816782 - `BetaManagedAgentsRetryStatusRetrying object`67836784 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.67856786 - `BetaManagedAgentsRetryStatusExhausted object`67876788 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.67896790 - `BetaManagedAgentsRetryStatusTerminal object`67916792 The session encountered a terminal error and will transition to `terminated` state.67936794 - `BetaManagedAgentsMCPAuthenticationFailedError object`67956796 Authentication to an MCP server failed.67976798 - `type: "mcp_authentication_failed_error"`67996800 - `mcp_server_name: string`68016802 Name of the MCP server that failed authentication.68036804 - `message: string`68056806 Human-readable error description.68076808 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`68096810 What the client should do next in response to this error.68116812 - `BetaManagedAgentsRetryStatusRetrying object`68136814 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.68156816 - `BetaManagedAgentsRetryStatusExhausted object`68176818 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.68196820 - `BetaManagedAgentsRetryStatusTerminal object`68216822 The session encountered a terminal error and will transition to `terminated` state.68236824 - `BetaManagedAgentsBillingError object`68256826 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.68276828 - `type: "billing_error"`68296830 - `message: string`68316832 Human-readable error description.68336834 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`68356836 What the client should do next in response to this error.68376838 - `BetaManagedAgentsRetryStatusRetrying object`68396840 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.68416842 - `BetaManagedAgentsRetryStatusExhausted object`68436844 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.68456846 - `BetaManagedAgentsRetryStatusTerminal object`68476848 The session encountered a terminal error and will transition to `terminated` state.68496850 - `BetaManagedAgentsCredentialHostUnreachableError object`68516852 An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.68536854 - `type: "credential_host_unreachable_error"`68556856 - `credential_id: string`68576858 ID of the affected credential.68596860 - `message: string`68616862 Human-readable error description.68636864 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`68656866 What the client should do next in response to this error.68676868 - `BetaManagedAgentsRetryStatusRetrying object`68696870 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.68716872 - `BetaManagedAgentsRetryStatusExhausted object`68736874 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.68756876 - `BetaManagedAgentsRetryStatusTerminal object`68776878 The session encountered a terminal error and will transition to `terminated` state.68796880 - `vault_id: string`68816882 ID of the vault containing the affected credential.68836884 - `processed_at: string`68856886 A timestamp in RFC 3339 format68876888 format: date-time68896890 - `BetaManagedAgentsSessionStatusRescheduledEvent object`68916892 Indicates the session is recovering from an error state and is rescheduled for execution.68936894 - `type: "session.status_rescheduled"`68956896 - `id: string`68976898 Unique identifier for this event.68996900 - `processed_at: string`69016902 A timestamp in RFC 3339 format69036904 format: date-time69056906 - `BetaManagedAgentsSessionStatusRunningEvent object`69076908 Indicates the session is actively running and the agent is working.69096910 - `type: "session.status_running"`69116912 - `id: string`69136914 Unique identifier for this event.69156916 - `processed_at: string`69176918 A timestamp in RFC 3339 format69196920 format: date-time69216922 - `BetaManagedAgentsSessionStatusIdleEvent object`69236924 Indicates the agent has paused and is awaiting user input.69256926 - `type: "session.status_idle"`69276928 - `id: string`69296930 Unique identifier for this event.69316932 - `processed_at: string`69336934 A timestamp in RFC 3339 format69356936 format: date-time69376938 - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`69396940 - `BetaManagedAgentsSessionEndTurn object`69416942 The agent completed its turn naturally and is ready for the next user message.69436944 - `type: "end_turn"`69456946 - `BetaManagedAgentsSessionRequiresAction object`69476948 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.69496950 - `type: "requires_action"`69516952 - `event_ids: array of string`69536954 The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.69556956 - `BetaManagedAgentsSessionRetriesExhausted object`69576958 The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.69596960 - `type: "retries_exhausted"`69616962 - `BetaManagedAgentsSessionBudgetReached object`69636964 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.69656966 - `type: "budget_reached"`69676968 - `BetaManagedAgentsSessionStatusTerminatedEvent object`69696970 Indicates the session has terminated, either due to an error or completion.69716972 - `type: "session.status_terminated"`69736974 - `id: string`69756976 Unique identifier for this event.69776978 - `processed_at: string`69796980 A timestamp in RFC 3339 format69816982 format: date-time69836984 - `BetaManagedAgentsSessionThreadCreatedEvent object`69856986 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.69876988 - `type: "session.thread_created"`69896990 - `id: string`69916992 Unique identifier for this event.69936994 - `agent_name: string`69956996 Name of the callable agent the thread runs.69976998 - `processed_at: string`69997000 A timestamp in RFC 3339 format70017002 format: date-time70037004 - `session_thread_id: string`70057006 Public `sthr_` ID of the newly created thread.70077008 - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`70097010 Emitted when an outcome evaluation cycle begins.70117012 - `type: "span.outcome_evaluation_start"`70137014 - `id: string`70157016 Unique identifier for this event.70177018 - `iteration: number`70197020 0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.70217022 format: int3270237024 - `outcome_id: string`70257026 The `outc_` ID of the outcome being evaluated.70277028 - `processed_at: string`70297030 A timestamp in RFC 3339 format70317032 format: date-time70337034 - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`70357036 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.70377038 - `type: "span.outcome_evaluation_end"`70397040 - `id: string`70417042 Unique identifier for this event.70437044 - `explanation: string`70457046 Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.70477048 - `iteration: number`70497050 0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.70517052 format: int3270537054 - `outcome_evaluation_start_id: string`70557056 The id of the corresponding `span.outcome_evaluation_start` event.70577058 - `outcome_id: string`70597060 The `outc_` ID of the outcome being evaluated.70617062 - `processed_at: string`70637064 A timestamp in RFC 3339 format70657066 format: date-time70677068 - `result: string`70697070 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.70717072 - `usage: BetaManagedAgentsSpanModelUsage`70737074 Token usage for a single model request.70757076 - `cache_creation_input_tokens: number`70777078 Tokens used to create prompt cache in this request.70797080 format: int3270817082 - `cache_read_input_tokens: number`70837084 Tokens read from prompt cache in this request.70857086 format: int3270877088 - `input_tokens: number`70897090 Input tokens consumed by this request.70917092 format: int3270937094 - `output_tokens: number`70957096 Output tokens generated by this request.70977098 format: int3270997100 - `speed: optional "standard" or "fast" or null`71017102 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.71037104 - `"standard"`71057106 - `"fast"`71077108 - `BetaManagedAgentsSpanModelRequestStartEvent object`71097110 Emitted when a model request is initiated by the agent.71117112 - `type: "span.model_request_start"`71137114 - `id: string`71157116 Unique identifier for this event.71177118 - `processed_at: string`71197120 A timestamp in RFC 3339 format71217122 format: date-time71237124 - `BetaManagedAgentsSpanModelRequestEndEvent object`71257126 Emitted when a model request completes.71277128 - `type: "span.model_request_end"`71297130 - `id: string`71317132 Unique identifier for this event.71337134 - `is_error: boolean or null`71357136 Whether the model request resulted in an error.71377138 - `model_request_start_id: string`71397140 The id of the corresponding `span.model_request_start` event.71417142 - `model_usage: BetaManagedAgentsSpanModelUsage`71437144 Token usage for a single model request.71457146 - `processed_at: string`71477148 A timestamp in RFC 3339 format71497150 format: date-time71517152 - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`71537154 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.71557156 - `type: "span.outcome_evaluation_ongoing"`71577158 - `id: string`71597160 Unique identifier for this event.71617162 - `iteration: number`71637164 0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.71657166 format: int3271677168 - `outcome_id: string`71697170 The `outc_` ID of the outcome being evaluated.71717172 - `processed_at: string`71737174 A timestamp in RFC 3339 format71757176 format: date-time71777178 - `BetaManagedAgentsUserDefineOutcomeEvent object`71797180 Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.71817182 - `type: "user.define_outcome"`71837184 - `id: string`71857186 Unique identifier for this event.71877188 - `description: string`71897190 What the agent should produce. Copied from the input event.71917192 - `max_iterations: number or null`71937194 Evaluate-then-revise cycles before giving up. Default 3, max 20.71957196 format: int3271977198 - `outcome_id: string`71997200 Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.72017202 - `processed_at: string`72037204 A timestamp in RFC 3339 format72057206 format: date-time72077208 - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`72097210 Rubric for grading the quality of an outcome.72117212 - `BetaManagedAgentsFileRubric object`72137214 Rubric referenced by a file uploaded via the Files API.72157216 - `type: "file"`72177218 - `file_id: string`72197220 ID of the rubric file.72217222 - `BetaManagedAgentsTextRubric object`72237224 Rubric content provided inline as text.72257226 - `type: "text"`72277228 - `content: string`72297230 Rubric content. Plain text or markdown — the grader treats it as freeform text.72317232 - `BetaManagedAgentsSessionDeletedEvent object`72337234 Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.72357236 - `type: "session.deleted"`72377238 - `id: string`72397240 Unique identifier for this event.72417242 - `processed_at: string`72437244 A timestamp in RFC 3339 format72457246 format: date-time72477248 - `BetaManagedAgentsSessionThreadStatusRunningEvent object`72497250 A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.72517252 - `type: "session.thread_status_running"`72537254 - `id: string`72557256 Unique identifier for this event.72577258 - `agent_name: string`72597260 Name of the agent the thread runs.72617262 - `processed_at: string`72637264 A timestamp in RFC 3339 format72657266 format: date-time72677268 - `session_thread_id: string`72697270 Public sthr_ ID of the thread that started running.72717272 - `BetaManagedAgentsSessionThreadStatusIdleEvent object`72737274 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.72757276 - `type: "session.thread_status_idle"`72777278 - `id: string`72797280 Unique identifier for this event.72817282 - `agent_name: string`72837284 Name of the agent the thread runs.72857286 - `processed_at: string`72877288 A timestamp in RFC 3339 format72897290 format: date-time72917292 - `session_thread_id: string`72937294 Public sthr_ ID of the thread that went idle.72957296 - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`72977298 - `BetaManagedAgentsSessionEndTurn object`72997300 The agent completed its turn naturally and is ready for the next user message.73017302 - `BetaManagedAgentsSessionRequiresAction object`73037304 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.73057306 - `BetaManagedAgentsSessionRetriesExhausted object`73077308 The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.73097310 - `BetaManagedAgentsSessionBudgetReached object`73117312 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.73137314 - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`73157316 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.73177318 - `type: "session.thread_status_terminated"`73197320 - `id: string`73217322 Unique identifier for this event.73237324 - `agent_name: string`73257326 Name of the agent the thread runs.73277328 - `processed_at: string`73297330 A timestamp in RFC 3339 format73317332 format: date-time73337334 - `session_thread_id: string`73357336 Public sthr_ ID of the thread that terminated.73377338 - `BetaManagedAgentsUserToolResultEvent object`73397340 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.73417342 - `type: "user.tool_result"`73437344 - `id: string`73457346 Unique identifier for this event.73477348 - `tool_use_id: string`73497350 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.73517352 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`73537354 The result content returned by the tool.73557356 - `BetaManagedAgentsTextBlock object`73577358 Regular text content.73597360 - `BetaManagedAgentsImageBlock object`73617362 Image content specified directly as base64 data or as a reference via a URL.73637364 - `BetaManagedAgentsDocumentBlock object`73657366 Document content, either specified directly as base64 data, as text, or as a reference via a URL.73677368 - `BetaManagedAgentsSearchResultBlock object`73697370 A block containing a web search result.73717372 - `is_error: optional boolean or null`73737374 Whether the tool execution resulted in an error.73757376 - `processed_at: optional string or null`73777378 A timestamp in RFC 3339 format73797380 format: date-time73817382 - `session_thread_id: optional string or null`73837384 Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.73857386 - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`73877388 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.73897390 - `type: "session.thread_status_rescheduled"`73917392 - `id: string`73937394 Unique identifier for this event.73957396 - `agent_name: string`73977398 Name of the agent the thread runs.73997400 - `processed_at: string`74017402 A timestamp in RFC 3339 format74037404 format: date-time74057406 - `session_thread_id: string`74077408 Public sthr_ ID of the thread that is retrying.74097410 - `BetaManagedAgentsSessionUpdatedEvent object`74117412 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.74137414 - `type: "session.updated"`74157416 - `id: string`74177418 Unique identifier for this event.74197420 - `processed_at: string`74217422 A timestamp in RFC 3339 format74237424 format: date-time74257426 - `agent: optional BetaManagedAgentsSessionAgent or null`74277428 Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.74297430 - `type: "agent"`74317432 - `id: string`74337434 - `description: string or null`74357436 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`74377438 - `type: "url"`74397440 - `name: string`74417442 - `url: string`74437444 - `model: BetaManagedAgentsModelConfig`74457446 Model identifier and configuration.74477448 - `id: BetaManagedAgentsModel`74497450 The model that will power your agent.74517452 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.74537454 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`74557456 The model that will power your agent.74577458 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.74597460 - `"claude-fable-5-1"`74617462 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows74637464 - `"claude-sonnet-5"`74657466 High-performance model for coding and agents74677468 - `"claude-fable-5"`74697470 Next generation of intelligence for the hardest knowledge work and coding problems74717472 - `"claude-opus-5"`74737474 Powerful intelligence for long-running agents and coding74757476 - `"claude-opus-4-8"`74777478 Powerful intelligence for long-running agents and coding74797480 - `"claude-opus-4-7"`74817482 Powerful intelligence for long-running agents and coding74837484 - `"claude-opus-4-6"`74857486 Powerful intelligence for long-running agents and coding74877488 - `"claude-sonnet-4-6"`74897490 Best combination of speed and intelligence74917492 - `"claude-haiku-4-5"`74937494 Fastest model with near-frontier intelligence74957496 - `"claude-haiku-4-5-20251001"`74977498 Fastest model with near-frontier intelligence74997500 - `"claude-opus-4-5"`75017502 Powerful intelligence for long-running agents and coding75037504 - `"claude-opus-4-5-20251101"`75057506 Powerful intelligence for long-running agents and coding75077508 - `"claude-sonnet-4-5"`75097510 High-performance model for agents and coding75117512 - `"claude-sonnet-4-5-20250929"`75137514 High-performance model for agents and coding75157516 - `string`75177518 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`75197520 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.75217522 - `BetaManagedAgentsEffortLow object`75237524 Low effort. Favors latency over reasoning depth.75257526 - `type: "low"`75277528 - `BetaManagedAgentsEffortMedium object`75297530 Medium effort. Balances latency and reasoning depth.75317532 - `type: "medium"`75337534 - `BetaManagedAgentsEffortHigh object`75357536 High effort. Favors reasoning depth.75377538 - `type: "high"`75397540 - `BetaManagedAgentsEffortXhigh object`75417542 Extra-high effort. Not all models accept this level.75437544 - `type: "xhigh"`75457546 - `BetaManagedAgentsEffortMax object`75477548 Maximum effort. Favors reasoning depth over latency.75497550 - `type: "max"`75517552 - `inference_geo: optional string`75537554 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.75557556 - `speed: optional "standard" or "fast"`75577558 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.75597560 - `"standard"`75617562 - `"fast"`75637564 - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`75657566 Resolved coordinator topology with full agent definitions for each roster member.75677568 - `type: "coordinator"`75697570 - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`75717572 Full `agent` definitions the coordinator may spawn as session threads.75737574 - `BetaManagedAgentsSessionThreadAgent object`75757576 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`.75777578 - `type: "agent"`75797580 - `id: string`75817582 - `description: string or null`75837584 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`75857586 - `type: "url"`75877588 - `name: string`75897590 - `url: string`75917592 - `model: BetaManagedAgentsModelConfig`75937594 Model identifier and configuration.75957596 - `name: string`75977598 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`75997600 - `BetaManagedAgentsAnthropicSkill object`76017602 A resolved Anthropic-managed skill.76037604 - `type: "anthropic"`76057606 - `skill_id: string`76077608 - `version: string`76097610 - `BetaManagedAgentsCustomSkill object`76117612 A resolved user-created custom skill.76137614 - `type: "custom"`76157616 - `skill_id: string`76177618 - `version: string`76197620 - `system: string or null`76217622 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`76237624 - `BetaManagedAgentsAgentToolset20260401 object`76257626 - `type: "agent_toolset_20260401"`76277628 - `configs: array of BetaManagedAgentsAgentToolConfig`76297630 - `BetaManagedAgentsBashToolConfig object`76317632 Configuration for the bash tool.76337634 - `type: "bash"`76357636 - `enabled: boolean`76377638 - `name: "bash"`76397640 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`76417642 Permission policy for tool execution.76437644 - `BetaManagedAgentsAlwaysAllowPolicy object`76457646 Tool calls are automatically approved without user confirmation.76477648 - `type: "always_allow"`76497650 - `BetaManagedAgentsAlwaysAskPolicy object`76517652 Tool calls require user confirmation before execution.76537654 - `type: "always_ask"`76557656 - `BetaManagedAgentsAutoPolicy object`76577658 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.76597660 - `type: "auto"`76617662 - `BetaManagedAgentsEditToolConfig object`76637664 Configuration for the edit tool.76657666 - `type: "edit"`76677668 - `enabled: boolean`76697670 - `name: "edit"`76717672 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`76737674 Permission policy for tool execution.76757676 - `BetaManagedAgentsAlwaysAllowPolicy object`76777678 Tool calls are automatically approved without user confirmation.76797680 - `BetaManagedAgentsAlwaysAskPolicy object`76817682 Tool calls require user confirmation before execution.76837684 - `BetaManagedAgentsAutoPolicy object`76857686 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.76877688 - `BetaManagedAgentsReadToolConfig object`76897690 Configuration for the read tool.76917692 - `type: "read"`76937694 - `enabled: boolean`76957696 - `name: "read"`76977698 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`76997700 Permission policy for tool execution.77017702 - `BetaManagedAgentsAlwaysAllowPolicy object`77037704 Tool calls are automatically approved without user confirmation.77057706 - `BetaManagedAgentsAlwaysAskPolicy object`77077708 Tool calls require user confirmation before execution.77097710 - `BetaManagedAgentsAutoPolicy object`77117712 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.77137714 - `BetaManagedAgentsWriteToolConfig object`77157716 Configuration for the write tool.77177718 - `type: "write"`77197720 - `enabled: boolean`77217722 - `name: "write"`77237724 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`77257726 Permission policy for tool execution.77277728 - `BetaManagedAgentsAlwaysAllowPolicy object`77297730 Tool calls are automatically approved without user confirmation.77317732 - `BetaManagedAgentsAlwaysAskPolicy object`77337734 Tool calls require user confirmation before execution.77357736 - `BetaManagedAgentsAutoPolicy object`77377738 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.77397740 - `BetaManagedAgentsGlobToolConfig object`77417742 Configuration for the glob tool.77437744 - `type: "glob"`77457746 - `enabled: boolean`77477748 - `name: "glob"`77497750 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`77517752 Permission policy for tool execution.77537754 - `BetaManagedAgentsAlwaysAllowPolicy object`77557756 Tool calls are automatically approved without user confirmation.77577758 - `BetaManagedAgentsAlwaysAskPolicy object`77597760 Tool calls require user confirmation before execution.77617762 - `BetaManagedAgentsAutoPolicy object`77637764 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.77657766 - `BetaManagedAgentsGrepToolConfig object`77677768 Configuration for the grep tool.77697770 - `type: "grep"`77717772 - `enabled: boolean`77737774 - `name: "grep"`77757776 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`77777778 Permission policy for tool execution.77797780 - `BetaManagedAgentsAlwaysAllowPolicy object`77817782 Tool calls are automatically approved without user confirmation.77837784 - `BetaManagedAgentsAlwaysAskPolicy object`77857786 Tool calls require user confirmation before execution.77877788 - `BetaManagedAgentsAutoPolicy object`77897790 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.77917792 - `BetaManagedAgentsWebFetchToolConfig object`77937794 Configuration for the web_fetch tool.77957796 - `type: "web_fetch"`77977798 - `enabled: boolean`77997800 - `name: "web_fetch"`78017802 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`78037804 Permission policy for tool execution.78057806 - `BetaManagedAgentsAlwaysAllowPolicy object`78077808 Tool calls are automatically approved without user confirmation.78097810 - `BetaManagedAgentsAlwaysAskPolicy object`78117812 Tool calls require user confirmation before execution.78137814 - `BetaManagedAgentsAutoPolicy object`78157816 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.78177818 - `allowed_domains: optional array of string`78197820 - `blocked_domains: optional array of string`78217822 - `max_content_tokens: optional number or null`78237824 format: int3278257826 - `BetaManagedAgentsWebSearchToolConfig object`78277828 Configuration for the web_search tool.78297830 - `type: "web_search"`78317832 - `enabled: boolean`78337834 - `name: "web_search"`78357836 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`78377838 Permission policy for tool execution.78397840 - `BetaManagedAgentsAlwaysAllowPolicy object`78417842 Tool calls are automatically approved without user confirmation.78437844 - `BetaManagedAgentsAlwaysAskPolicy object`78457846 Tool calls require user confirmation before execution.78477848 - `BetaManagedAgentsAutoPolicy object`78497850 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.78517852 - `allowed_domains: optional array of string`78537854 - `blocked_domains: optional array of string`78557856 - `user_location: optional BetaManagedAgentsUserLocation or null`78577858 Approximate user location for search result localization.78597860 - `type: "approximate"`78617862 Location precision. Only "approximate" is supported.78637864 - `city: optional string or null`78657866 City name.78677868 minLength: 1, maxLength: 25578697870 - `country: optional string or null`78717872 Two-letter ISO 3166-1 country code, uppercase.78737874 - `region: optional string or null`78757876 Region or state name.78777878 minLength: 1, maxLength: 25578797880 - `timezone: optional string or null`78817882 IANA timezone identifier, e.g. "America/Los_Angeles".78837884 minLength: 1, maxLength: 25578857886 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`78877888 Resolved default configuration for agent tools.78897890 - `enabled: boolean`78917892 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`78937894 Permission policy for tool execution.78957896 - `BetaManagedAgentsAlwaysAllowPolicy object`78977898 Tool calls are automatically approved without user confirmation.78997900 - `BetaManagedAgentsAlwaysAskPolicy object`79017902 Tool calls require user confirmation before execution.79037904 - `BetaManagedAgentsAutoPolicy object`79057906 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.79077908 - `BetaManagedAgentsMCPToolset object`79097910 - `type: "mcp_toolset"`79117912 - `configs: array of BetaManagedAgentsMCPToolConfig`79137914 - `enabled: boolean`79157916 - `name: string`79177918 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`79197920 Permission policy for tool execution.79217922 - `BetaManagedAgentsAlwaysAllowPolicy object`79237924 Tool calls are automatically approved without user confirmation.79257926 - `BetaManagedAgentsAlwaysAskPolicy object`79277928 Tool calls require user confirmation before execution.79297930 - `BetaManagedAgentsAutoPolicy object`79317932 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.79337934 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`79357936 Resolved default configuration for all tools from an MCP server.79377938 - `enabled: boolean`79397940 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`79417942 Permission policy for tool execution.79437944 - `BetaManagedAgentsAlwaysAllowPolicy object`79457946 Tool calls are automatically approved without user confirmation.79477948 - `BetaManagedAgentsAlwaysAskPolicy object`79497950 Tool calls require user confirmation before execution.79517952 - `BetaManagedAgentsAutoPolicy object`79537954 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.79557956 - `mcp_server_name: string`79577958 - `BetaManagedAgentsCustomTool object`79597960 A custom tool as returned in API responses.79617962 - `type: "custom"`79637964 - `description: string`79657966 - `input_schema: BetaManagedAgentsCustomToolInputSchema`79677968 JSON Schema for custom tool input parameters.79697970 - `type: "object"`79717972 - `properties: optional map[unknown] or null`79737974 - `required: optional array of string or null`79757976 - `name: string`79777978 - `version: number`79797980 format: int3279817982 - `BetaManagedAgentsAdvisor object`79837984 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.79857986 - `type: "advisor"`79877988 - `model: string`79897990 The advisor model id.79917992 - `name: string`79937994 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`79957996 - `BetaManagedAgentsAnthropicSkill object`79977998 A resolved Anthropic-managed skill.79998000 - `BetaManagedAgentsCustomSkill object`80018002 A resolved user-created custom skill.80038004 - `system: string or null`80058006 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`80078008 - `BetaManagedAgentsAgentToolset20260401 object`80098010 - `BetaManagedAgentsMCPToolset object`80118012 - `BetaManagedAgentsCustomTool object`80138014 A custom tool as returned in API responses.80158016 - `version: number`80178018 format: int3280198020 - `budget: optional BetaManagedAgentsBudgetLimit or null`80218022 A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.80238024 - `type: "limit"`80258026 - `max_list_cost: BetaMonetaryAmount`80278028 A monetary amount in a specific currency.80298030 - `amount: string`80318032 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.80338034 - `currency: BetaCurrency`80358036 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.80378038 - `metadata: optional map[string]`80398040 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.80418042 - `title: optional string or null`80438044 The session's new title. Present only when the update changed it.80458046 - `BetaManagedAgentsSystemMessageEvent object`80478048 A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.80498050 - `type: "system.message"`80518052 - `id: string`80538054 Unique identifier for this event.80558056 - `content: array of BetaManagedAgentsSystemContentBlock`80578058 System content blocks. Text-only.80598060 - `type: "text"`80618062 - `text: string`80638064 The text content.80658066 minLength: 180678068 - `processed_at: optional string or null`80698070 A timestamp in RFC 3339 format80718072 format: date-time80738074 - `BetaManagedAgentsSessionUsageEvent object`80758076 Periodic snapshot of the session's cumulative usage and tracked list cost.80778078 - `type: "session.usage"`80798080 - `id: string`80818082 Unique identifier for this event.80838084 - `processed_at: string`80858086 A timestamp in RFC 3339 format80878088 format: date-time80898090 - `usage: BetaManagedAgentsSessionUsageSnapshot`80918092 Point-in-time snapshot of a session's cumulative usage.80938094 - `active_seconds: optional number`80958096 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.80978098 format: double80998100 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`81018102 Prompt-cache creation token usage broken down by cache lifetime.81038104 - `ephemeral_1h_input_tokens: optional number`81058106 Tokens used to create 1-hour ephemeral cache entries.81078108 format: int3281098110 - `ephemeral_5m_input_tokens: optional number`81118112 Tokens used to create 5-minute ephemeral cache entries.81138114 format: int3281158116 - `cache_read_input_tokens: optional number`81178118 Total tokens read from prompt cache.81198120 format: int3281218122 - `input_tokens: optional number`81238124 Total input tokens consumed across all turns.81258126 format: int3281278128 - `list_cost: optional BetaMonetaryAmount`81298130 A monetary amount in a specific currency.81318132 - `output_tokens: optional number`81338134 Total output tokens generated across all turns.81358136 format: int3281378138 - `server_tool_use: optional BetaManagedAgentsServerToolUsage`81398140 Cumulative count of server-executed tool invocations, broken down by tool.81418142 - `web_fetch_requests: optional number`81438144 Number of server-executed web fetch requests.81458146 format: int3281478148 - `web_search_requests: optional number`81498150 Number of server-executed web search requests.81518152 format: int3281538154 - `budget: optional BetaManagedAgentsBudgetLimit or null`81558156 A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.81578158- `next_page: optional string or null`81598160 Opaque cursor for the next page. Null when no more results.81618162#### Example81638164```bash8165curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID/events \8166 -H 'anthropic-version: 2023-06-01' \8167 -H 'anthropic-beta: managed-agents-2026-04-01' \8168 -H "X-Api-Key: $ANTHROPIC_API_KEY"8169```81708171##### Response (200)81728173```json8174{8175 "data": [8176 {8177 "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",8178 "content": [8179 {8180 "text": "Where is my order #1234?",8181 "type": "text"8182 }8183 ],8184 "type": "user.message",8185 "processed_at": "2026-03-15T10:00:00Z"8186 }8187 ],8188 "next_page": "next_page"8189}8190```81918192### Stream Session Thread Events81938194**GET** `/v1/sessions/{session_id}/threads/{thread_id}/stream`81958196Stream Session Thread Events81978198#### Path parameters81998200- `session_id: string`82018202- `thread_id: string`82038204#### Query parameters82058206- `event_deltas: optional array of BetaManagedAgentsDeltaType`82078208 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.82098210 - `"agent.message"`82118212 - `"agent.thinking"`82138214#### Headers82158216- `"anthropic-beta": optional array of AnthropicBeta`82178218 Optional header to specify the beta version(s) you want to use.82198220 - `string`82218222 - `"message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more`82238224 - `"message-batches-2024-09-24"`82258226 - `"prompt-caching-2024-07-31"`82278228 - `"computer-use-2024-10-22"`82298230 - `"computer-use-2025-01-24"`82318232 - `"pdfs-2024-09-25"`82338234 - `"token-counting-2024-11-01"`82358236 - `"token-efficient-tools-2025-02-19"`82378238 - `"output-128k-2025-02-19"`82398240 - `"files-api-2025-04-14"`82418242 - `"mcp-client-2025-04-04"`82438244 - `"mcp-client-2025-11-20"`82458246 - `"dev-full-thinking-2025-05-14"`82478248 - `"interleaved-thinking-2025-05-14"`82498250 - `"code-execution-2025-05-22"`82518252 - `"extended-cache-ttl-2025-04-11"`82538254 - `"context-1m-2025-08-07"`82558256 - `"context-management-2025-06-27"`82578258 - `"model-context-window-exceeded-2025-08-26"`82598260 - `"skills-2025-10-02"`82618262 - `"fast-mode-2026-02-01"`82638264 - `"output-300k-2026-03-24"`82658266 - `"user-profiles-2026-03-24"`82678268 - `"user-profiles-2026-08-18"`82698270 - `"user-profiles-2026-09-04"`82718272 - `"advisor-tool-2026-03-01"`82738274 - `"managed-agents-2026-04-01"`82758276 - `"cache-diagnosis-2026-04-07"`82778278 - `"dreaming-2026-04-21"`82798280 - `"thinking-token-count-2026-05-13"`82818282 - `"server-side-fallback-2026-06-01"`82838284 - `"server-side-fallback-2026-07-01"`82858286 - `"fallback-credit-2026-06-01"`82878288 - `"fallback-credit-2026-07-01"`82898290 - `"agent-memory-2026-07-22"`82918292 - `"mid-conversation-tool-changes-2026-07-01"`82938294 - `"compact-2026-01-12"`82958296 - `"computer-use-2025-11-24"`82978298 - `"mcp-tunnels-2026-06-22"`82998300 - `"structured-outputs-2025-11-13"`83018302 - `"task-budgets-2026-03-13"`83038304 - `"thinking-display-updates-2026-08-18"`83058306 - `"ce-user-management-2026-07-13"`83078308 - `"mid-conversation-output-config-2026-07-01"`83098310 - `"thinking-binding-controls-2026-08-01"`83118312 - `"mid-conversation-system-clear-at-2026-08-21"`83138314 - `"compact-2026-09-04"`83158316- `"anthropic-workspace-id": optional string`83178318#### Returns83198320- `BetaManagedAgentsStreamSessionThreadEvents = BetaManagedAgentsUserMessageEvent or BetaManagedAgentsUserInterruptEvent or BetaManagedAgentsUserToolConfirmationEvent or 34 more`83218322 Server-sent event in a single thread's stream.83238324 - `BetaManagedAgentsUserMessageEvent object`83258326 A user message event in the session conversation.83278328 - `type: "user.message"`83298330 - `id: string`83318332 Unique identifier for this event.83338334 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`83358336 Array of content blocks comprising the user message.83378338 - `BetaManagedAgentsTextBlock object`83398340 Regular text content.83418342 - `type: "text"`83438344 - `text: string`83458346 The text content.83478348 minLength: 183498350 - `BetaManagedAgentsImageBlock object`83518352 Image content specified directly as base64 data or as a reference via a URL.83538354 - `type: "image"`83558356 - `source: BetaManagedAgentsBase64ImageSource or BetaManagedAgentsURLImageSource or BetaManagedAgentsFileImageSource`83578358 Union type for image source variants.83598360 - `BetaManagedAgentsBase64ImageSource object`83618362 Base64-encoded image data.83638364 - `type: "base64"`83658366 - `data: string`83678368 Base64-encoded image data.83698370 minLength: 183718372 - `media_type: string`83738374 MIME type of the image (e.g., "image/png", "image/jpeg", "image/gif", "image/webp").83758376 minLength: 183778378 - `BetaManagedAgentsURLImageSource object`83798380 Image referenced by URL.83818382 - `type: "url"`83838384 - `url: string`83858386 URL of the image to fetch.83878388 minLength: 183898390 - `BetaManagedAgentsFileImageSource object`83918392 Image referenced by file ID.83938394 - `type: "file"`83958396 - `file_id: string`83978398 ID of a previously uploaded file.83998400 minLength: 184018402 - `BetaManagedAgentsDocumentBlock object`84038404 Document content, either specified directly as base64 data, as text, or as a reference via a URL.84058406 - `type: "document"`84078408 - `source: BetaManagedAgentsBase64DocumentSource or BetaManagedAgentsPlainTextDocumentSource or BetaManagedAgentsURLDocumentSource or BetaManagedAgentsFileDocumentSource`84098410 Union type for document source variants.84118412 - `BetaManagedAgentsBase64DocumentSource object`84138414 Base64-encoded document data.84158416 - `type: "base64"`84178418 - `data: string`84198420 Base64-encoded document data.84218422 minLength: 184238424 - `media_type: string`84258426 MIME type of the document (e.g., "application/pdf").84278428 minLength: 184298430 - `BetaManagedAgentsPlainTextDocumentSource object`84318432 Plain text document content.84338434 - `type: "text"`84358436 - `data: string`84378438 The plain text content.84398440 minLength: 184418442 - `media_type: "text/plain"`84438444 MIME type of the text content. Must be "text/plain".84458446 - `BetaManagedAgentsURLDocumentSource object`84478448 Document referenced by URL.84498450 - `type: "url"`84518452 - `url: string`84538454 URL of the document to fetch.84558456 minLength: 184578458 - `BetaManagedAgentsFileDocumentSource object`84598460 Document referenced by file ID.84618462 - `type: "file"`84638464 - `file_id: string`84658466 ID of a previously uploaded file.84678468 minLength: 184698470 - `context: optional string or null`84718472 Additional context about the document for the model.84738474 - `title: optional string or null`84758476 The title of the document.84778478 - `BetaManagedAgentsRedactedBlock object`84798480 Placeholder for content withheld by Anthropic model policy.84818482 - `type: "redacted"`84838484 - `processed_at: optional string or null`84858486 A timestamp in RFC 3339 format84878488 format: date-time84898490 - `BetaManagedAgentsUserInterruptEvent object`84918492 An interrupt event that pauses agent execution and returns control to the user.84938494 - `type: "user.interrupt"`84958496 - `id: string`84978498 Unique identifier for this event.84998500 - `processed_at: optional string or null`85018502 A timestamp in RFC 3339 format85038504 format: date-time85058506 - `session_thread_id: optional string or null`85078508 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.85098510 - `BetaManagedAgentsUserToolConfirmationEvent object`85118512 A tool confirmation event that approves or denies a pending tool execution.85138514 - `type: "user.tool_confirmation"`85158516 - `id: string`85178518 Unique identifier for this event.85198520 - `result: "allow" or "deny"`85218522 UserToolConfirmationResult enum85238524 - `"allow"`85258526 - `"deny"`85278528 - `tool_use_id: string`85298530 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.85318532 - `deny_message: optional string or null`85338534 Optional message providing context for a 'deny' decision. Only allowed when result is 'deny'.85358536 maxLength: 1000085378538 - `processed_at: optional string or null`85398540 A timestamp in RFC 3339 format85418542 format: date-time85438544 - `session_thread_id: optional string or null`85458546 Set by the server to the subagent thread this confirmation was routed to. Omitted when it was routed to the primary thread.85478548 - `BetaManagedAgentsUserCustomToolResultEvent object`85498550 Event sent by the client providing the result of a custom tool execution.85518552 - `type: "user.custom_tool_result"`85538554 - `id: string`85558556 Unique identifier for this event.85578558 - `custom_tool_use_id: string`85598560 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.85618562 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`85638564 The result content returned by the tool.85658566 - `BetaManagedAgentsTextBlock object`85678568 Regular text content.85698570 - `BetaManagedAgentsImageBlock object`85718572 Image content specified directly as base64 data or as a reference via a URL.85738574 - `BetaManagedAgentsDocumentBlock object`85758576 Document content, either specified directly as base64 data, as text, or as a reference via a URL.85778578 - `BetaManagedAgentsSearchResultBlock object`85798580 A block containing a web search result.85818582 - `type: "search_result"`85838584 - `citations: BetaManagedAgentsSearchResultCitations`85858586 Citation settings for a search result.85878588 - `enabled: boolean`85898590 Whether citations are enabled for this search result.85918592 - `content: array of BetaManagedAgentsSearchResultContent`85938594 Array of text content blocks from the search result.85958596 - `type: "text"`85978598 - `text: string`85998600 The text content.86018602 minLength: 186038604 - `source: string`86058606 The URL source of the search result.86078608 minLength: 186098610 - `title: string`86118612 The title of the search result.86138614 minLength: 186158616 - `is_error: optional boolean or null`86178618 Whether the tool execution resulted in an error.86198620 - `processed_at: optional string or null`86218622 A timestamp in RFC 3339 format86238624 format: date-time86258626 - `session_thread_id: optional string or null`86278628 Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.86298630 - `BetaManagedAgentsAgentCustomToolUseEvent object`86318632 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.86338634 - `type: "agent.custom_tool_use"`86358636 - `id: string`86378638 Unique identifier for this event.86398640 - `input: map[unknown]`86418642 Input parameters for the tool call.86438644 - `name: string`86458646 Name of the custom tool being called.86478648 - `processed_at: string`86498650 A timestamp in RFC 3339 format86518652 format: date-time86538654 - `session_thread_id: optional string or null`86558656 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.86578658 - `BetaManagedAgentsAgentMessageEvent object`86598660 An agent response event in the session conversation.86618662 - `type: "agent.message"`86638664 - `id: string`86658666 Unique identifier for this event.86678668 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsRedactedBlock`86698670 Array of text blocks comprising the agent response.86718672 - `BetaManagedAgentsTextBlock object`86738674 Regular text content.86758676 - `BetaManagedAgentsRedactedBlock object`86778678 Placeholder for content withheld by Anthropic model policy.86798680 - `processed_at: string`86818682 A timestamp in RFC 3339 format86838684 format: date-time86858686 - `BetaManagedAgentsAgentThinkingEvent object`86878688 Indicates the agent is making forward progress via extended thinking. A progress signal, not a content carrier.86898690 - `type: "agent.thinking"`86918692 - `id: string`86938694 Unique identifier for this event.86958696 - `processed_at: string`86978698 A timestamp in RFC 3339 format86998700 format: date-time87018702 - `BetaManagedAgentsAgentMCPToolUseEvent object`87038704 Event emitted when the agent invokes a tool provided by an MCP server.87058706 - `type: "agent.mcp_tool_use"`87078708 - `id: string`87098710 Unique identifier for this event.87118712 - `input: map[unknown]`87138714 Input parameters for the tool call.87158716 - `mcp_server_name: string`87178718 Name of the MCP server providing the tool.87198720 - `name: string`87218722 Name of the MCP tool being used.87238724 - `processed_at: string`87258726 A timestamp in RFC 3339 format87278728 format: date-time87298730 - `evaluated_permission: optional "allow" or "ask" or "deny"`87318732 AgentEvaluatedPermission enum87338734 - `"allow"`87358736 - `"ask"`87378738 - `"deny"`87398740 - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`87418742 Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.87438744 - `BetaManagedAgentsAgentToolEvaluationAlwaysAllow object`87458746 The resolved permission_policy was always_allow; accompanies evaluated_permission "allow".87478748 - `type: "always_allow"`87498750 - `BetaManagedAgentsAgentToolEvaluationAlwaysAsk object`87518752 The resolved permission_policy was always_ask; accompanies evaluated_permission "ask".87538754 - `type: "always_ask"`87558756 - `BetaManagedAgentsAgentToolEvaluationAuto object`87578758 The resolved permission_policy was auto: the server judged this invocation individually.87598760 - `type: "auto"`87618762 - `evaluated_permission: BetaManagedAgentsAgentAutoEvaluatedPermission`87638764 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.87658766 - `BetaManagedAgentsAgentAutoEvaluatedPermissionAllow object`87678768 The server judged the invocation safe to execute without client approval.87698770 - `type: "allow"`87718772 - `BetaManagedAgentsAgentAutoEvaluatedPermissionAsk object`87738774 The server reached no judgement; the invocation is held for client approval.87758776 - `type: "ask"`87778778 - `reason_code: string`87798780 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.87818782 maxLength: 6487838784 - `BetaManagedAgentsAgentAutoEvaluatedPermissionDeny object`87858786 The server judged the invocation high-risk; it does not execute and a synthetic error tool result is appended.87878788 - `type: "deny"`87898790 - `reason_code: string`87918792 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.87938794 maxLength: 6487958796 - `session_thread_id: optional string or null`87978798 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.87998800 - `BetaManagedAgentsAgentMCPToolResultEvent object`88018802 Event representing the result of an MCP tool execution.88038804 - `type: "agent.mcp_tool_result"`88058806 - `id: string`88078808 Unique identifier for this event.88098810 - `mcp_tool_use_id: string`88118812 The id of the `agent.mcp_tool_use` event this result corresponds to.88138814 - `processed_at: string`88158816 A timestamp in RFC 3339 format88178818 format: date-time88198820 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`88218822 The result content returned by the tool.88238824 - `BetaManagedAgentsTextBlock object`88258826 Regular text content.88278828 - `BetaManagedAgentsImageBlock object`88298830 Image content specified directly as base64 data or as a reference via a URL.88318832 - `BetaManagedAgentsDocumentBlock object`88338834 Document content, either specified directly as base64 data, as text, or as a reference via a URL.88358836 - `BetaManagedAgentsSearchResultBlock object`88378838 A block containing a web search result.88398840 - `is_error: optional boolean or null`88418842 Whether the tool execution resulted in an error.88438844 - `BetaManagedAgentsAgentToolUseEvent object`88458846 Event emitted when the agent invokes a built-in agent tool.88478848 - `type: "agent.tool_use"`88498850 - `id: string`88518852 Unique identifier for this event.88538854 - `input: map[unknown]`88558856 Input parameters for the tool call.88578858 - `name: string`88598860 Name of the agent tool being used.88618862 - `processed_at: string`88638864 A timestamp in RFC 3339 format88658866 format: date-time88678868 - `evaluated_permission: optional "allow" or "ask" or "deny"`88698870 AgentEvaluatedPermission enum88718872 - `"allow"`88738874 - `"ask"`88758876 - `"deny"`88778878 - `evaluation: optional BetaManagedAgentsAgentToolEvaluation`88798880 Names the resolved permission_policy that produced evaluated_permission, and under auto carries the judgement. Open union: clients must tolerate unknown variants.88818882 - `session_thread_id: optional string or null`88838884 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.88858886 - `BetaManagedAgentsAgentToolResultEvent object`88878888 Event representing the result of an agent tool execution.88898890 - `type: "agent.tool_result"`88918892 - `id: string`88938894 Unique identifier for this event.88958896 - `processed_at: string`88978898 A timestamp in RFC 3339 format88998900 format: date-time89018902 - `tool_use_id: string`89038904 The id of the `agent.tool_use` event this result corresponds to.89058906 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`89078908 The result content returned by the tool.89098910 - `BetaManagedAgentsTextBlock object`89118912 Regular text content.89138914 - `BetaManagedAgentsImageBlock object`89158916 Image content specified directly as base64 data or as a reference via a URL.89178918 - `BetaManagedAgentsDocumentBlock object`89198920 Document content, either specified directly as base64 data, as text, or as a reference via a URL.89218922 - `BetaManagedAgentsSearchResultBlock object`89238924 A block containing a web search result.89258926 - `is_error: optional boolean or null`89278928 Whether the tool execution resulted in an error.89298930 - `BetaManagedAgentsAgentThreadMessageReceivedEvent object`89318932 Delivery event written to the target thread's input stream when an agent-to-agent message arrives.89338934 - `type: "agent.thread_message_received"`89358936 - `id: string`89378938 Unique identifier for this event.89398940 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`89418942 Message content blocks.89438944 - `BetaManagedAgentsTextBlock object`89458946 Regular text content.89478948 - `BetaManagedAgentsImageBlock object`89498950 Image content specified directly as base64 data or as a reference via a URL.89518952 - `BetaManagedAgentsDocumentBlock object`89538954 Document content, either specified directly as base64 data, as text, or as a reference via a URL.89558956 - `BetaManagedAgentsRedactedBlock object`89578958 Placeholder for content withheld by Anthropic model policy.89598960 - `from_session_thread_id: string`89618962 Public `sthr_` ID of the thread that sent the message.89638964 - `processed_at: string`89658966 A timestamp in RFC 3339 format89678968 format: date-time89698970 - `from_agent_name: optional string or null`89718972 Name of the callable agent this message came from. Absent when received from the primary agent.89738974 - `BetaManagedAgentsAgentThreadMessageSentEvent object`89758976 Observability event emitted to the sender's output stream when an agent-to-agent message is sent.89778978 - `type: "agent.thread_message_sent"`89798980 - `id: string`89818982 Unique identifier for this event.89838984 - `content: array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsRedactedBlock`89858986 Message content blocks.89878988 - `BetaManagedAgentsTextBlock object`89898990 Regular text content.89918992 - `BetaManagedAgentsImageBlock object`89938994 Image content specified directly as base64 data or as a reference via a URL.89958996 - `BetaManagedAgentsDocumentBlock object`89978998 Document content, either specified directly as base64 data, as text, or as a reference via a URL.89999000 - `BetaManagedAgentsRedactedBlock object`90019002 Placeholder for content withheld by Anthropic model policy.90039004 - `processed_at: string`90059006 A timestamp in RFC 3339 format90079008 format: date-time90099010 - `to_session_thread_id: string`90119012 Public `sthr_` ID of the thread the message was sent to.90139014 - `to_agent_name: optional string or null`90159016 Name of the callable agent this message was sent to. Absent when sent to the primary agent.90179018 - `BetaManagedAgentsAgentThreadContextCompactedEvent object`90199020 Indicates that context compaction (summarization) occurred during the session.90219022 - `type: "agent.thread_context_compacted"`90239024 - `id: string`90259026 Unique identifier for this event.90279028 - `processed_at: string`90299030 A timestamp in RFC 3339 format90319032 format: date-time90339034 - `BetaManagedAgentsSessionErrorEvent object`90359036 An error event indicating a problem occurred during session execution.90379038 - `type: "session.error"`90399040 - `id: string`90419042 Unique identifier for this event.90439044 - `error: BetaManagedAgentsUnknownError or BetaManagedAgentsModelOverloadedError or BetaManagedAgentsModelRateLimitedError or 5 more`90459046 - `BetaManagedAgentsUnknownError object`90479048 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.90499050 - `type: "unknown_error"`90519052 - `message: string`90539054 Human-readable error description.90559056 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`90579058 What the client should do next in response to this error.90599060 - `BetaManagedAgentsRetryStatusRetrying object`90619062 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.90639064 - `type: "retrying"`90659066 - `BetaManagedAgentsRetryStatusExhausted object`90679068 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.90699070 - `type: "exhausted"`90719072 - `BetaManagedAgentsRetryStatusTerminal object`90739074 The session encountered a terminal error and will transition to `terminated` state.90759076 - `type: "terminal"`90779078 - `BetaManagedAgentsModelOverloadedError object`90799080 The model is currently overloaded. Emitted after automatic retries are exhausted.90819082 - `type: "model_overloaded_error"`90839084 - `message: string`90859086 Human-readable error description.90879088 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`90899090 What the client should do next in response to this error.90919092 - `BetaManagedAgentsRetryStatusRetrying object`90939094 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.90959096 - `BetaManagedAgentsRetryStatusExhausted object`90979098 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.90999100 - `BetaManagedAgentsRetryStatusTerminal object`91019102 The session encountered a terminal error and will transition to `terminated` state.91039104 - `BetaManagedAgentsModelRateLimitedError object`91059106 The model request was rate-limited.91079108 - `type: "model_rate_limited_error"`91099110 - `message: string`91119112 Human-readable error description.91139114 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`91159116 What the client should do next in response to this error.91179118 - `BetaManagedAgentsRetryStatusRetrying object`91199120 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.91219122 - `BetaManagedAgentsRetryStatusExhausted object`91239124 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.91259126 - `BetaManagedAgentsRetryStatusTerminal object`91279128 The session encountered a terminal error and will transition to `terminated` state.91299130 - `BetaManagedAgentsModelRequestFailedError object`91319132 A model request failed for a reason other than overload or rate-limiting.91339134 - `type: "model_request_failed_error"`91359136 - `message: string`91379138 Human-readable error description.91399140 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`91419142 What the client should do next in response to this error.91439144 - `BetaManagedAgentsRetryStatusRetrying object`91459146 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.91479148 - `BetaManagedAgentsRetryStatusExhausted object`91499150 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.91519152 - `BetaManagedAgentsRetryStatusTerminal object`91539154 The session encountered a terminal error and will transition to `terminated` state.91559156 - `BetaManagedAgentsMCPConnectionFailedError object`91579158 Failed to connect to an MCP server.91599160 - `type: "mcp_connection_failed_error"`91619162 - `mcp_server_name: string`91639164 Name of the MCP server that failed to connect.91659166 - `message: string`91679168 Human-readable error description.91699170 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`91719172 What the client should do next in response to this error.91739174 - `BetaManagedAgentsRetryStatusRetrying object`91759176 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.91779178 - `BetaManagedAgentsRetryStatusExhausted object`91799180 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.91819182 - `BetaManagedAgentsRetryStatusTerminal object`91839184 The session encountered a terminal error and will transition to `terminated` state.91859186 - `BetaManagedAgentsMCPAuthenticationFailedError object`91879188 Authentication to an MCP server failed.91899190 - `type: "mcp_authentication_failed_error"`91919192 - `mcp_server_name: string`91939194 Name of the MCP server that failed authentication.91959196 - `message: string`91979198 Human-readable error description.91999200 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`92019202 What the client should do next in response to this error.92039204 - `BetaManagedAgentsRetryStatusRetrying object`92059206 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.92079208 - `BetaManagedAgentsRetryStatusExhausted object`92099210 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.92119212 - `BetaManagedAgentsRetryStatusTerminal object`92139214 The session encountered a terminal error and will transition to `terminated` state.92159216 - `BetaManagedAgentsBillingError object`92179218 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.92199220 - `type: "billing_error"`92219222 - `message: string`92239224 Human-readable error description.92259226 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`92279228 What the client should do next in response to this error.92299230 - `BetaManagedAgentsRetryStatusRetrying object`92319232 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.92339234 - `BetaManagedAgentsRetryStatusExhausted object`92359236 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.92379238 - `BetaManagedAgentsRetryStatusTerminal object`92399240 The session encountered a terminal error and will transition to `terminated` state.92419242 - `BetaManagedAgentsCredentialHostUnreachableError object`92439244 An `environment_variable` credential's `auth.networking.allowed_hosts` includes a host the environment's network policy does not permit.92459246 - `type: "credential_host_unreachable_error"`92479248 - `credential_id: string`92499250 ID of the affected credential.92519252 - `message: string`92539254 Human-readable error description.92559256 - `retry_status: BetaManagedAgentsRetryStatusRetrying or BetaManagedAgentsRetryStatusExhausted or BetaManagedAgentsRetryStatusTerminal`92579258 What the client should do next in response to this error.92599260 - `BetaManagedAgentsRetryStatusRetrying object`92619262 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.92639264 - `BetaManagedAgentsRetryStatusExhausted object`92659266 This turn is dead; queued inputs are flushed and the session returns to idle. Client may send a new prompt.92679268 - `BetaManagedAgentsRetryStatusTerminal object`92699270 The session encountered a terminal error and will transition to `terminated` state.92719272 - `vault_id: string`92739274 ID of the vault containing the affected credential.92759276 - `processed_at: string`92779278 A timestamp in RFC 3339 format92799280 format: date-time92819282 - `BetaManagedAgentsSessionStatusRescheduledEvent object`92839284 Indicates the session is recovering from an error state and is rescheduled for execution.92859286 - `type: "session.status_rescheduled"`92879288 - `id: string`92899290 Unique identifier for this event.92919292 - `processed_at: string`92939294 A timestamp in RFC 3339 format92959296 format: date-time92979298 - `BetaManagedAgentsSessionStatusRunningEvent object`92999300 Indicates the session is actively running and the agent is working.93019302 - `type: "session.status_running"`93039304 - `id: string`93059306 Unique identifier for this event.93079308 - `processed_at: string`93099310 A timestamp in RFC 3339 format93119312 format: date-time93139314 - `BetaManagedAgentsSessionStatusIdleEvent object`93159316 Indicates the agent has paused and is awaiting user input.93179318 - `type: "session.status_idle"`93199320 - `id: string`93219322 Unique identifier for this event.93239324 - `processed_at: string`93259326 A timestamp in RFC 3339 format93279328 format: date-time93299330 - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`93319332 - `BetaManagedAgentsSessionEndTurn object`93339334 The agent completed its turn naturally and is ready for the next user message.93359336 - `type: "end_turn"`93379338 - `BetaManagedAgentsSessionRequiresAction object`93399340 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.93419342 - `type: "requires_action"`93439344 - `event_ids: array of string`93459346 The ids of events the agent is blocked on. Resolving fewer than all re-emits `session.status_idle` with the remainder.93479348 - `BetaManagedAgentsSessionRetriesExhausted object`93499350 The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.93519352 - `type: "retries_exhausted"`93539354 - `BetaManagedAgentsSessionBudgetReached object`93559356 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.93579358 - `type: "budget_reached"`93599360 - `BetaManagedAgentsSessionStatusTerminatedEvent object`93619362 Indicates the session has terminated, either due to an error or completion.93639364 - `type: "session.status_terminated"`93659366 - `id: string`93679368 Unique identifier for this event.93699370 - `processed_at: string`93719372 A timestamp in RFC 3339 format93739374 format: date-time93759376 - `BetaManagedAgentsSessionThreadCreatedEvent object`93779378 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.93799380 - `type: "session.thread_created"`93819382 - `id: string`93839384 Unique identifier for this event.93859386 - `agent_name: string`93879388 Name of the callable agent the thread runs.93899390 - `processed_at: string`93919392 A timestamp in RFC 3339 format93939394 format: date-time93959396 - `session_thread_id: string`93979398 Public `sthr_` ID of the newly created thread.93999400 - `BetaManagedAgentsSpanOutcomeEvaluationStartEvent object`94019402 Emitted when an outcome evaluation cycle begins.94039404 - `type: "span.outcome_evaluation_start"`94059406 - `id: string`94079408 Unique identifier for this event.94099410 - `iteration: number`94119412 0-indexed revision cycle. 0 is the first evaluation; 1 is the re-evaluation after the first revision; etc.94139414 format: int3294159416 - `outcome_id: string`94179418 The `outc_` ID of the outcome being evaluated.94199420 - `processed_at: string`94219422 A timestamp in RFC 3339 format94239424 format: date-time94259426 - `BetaManagedAgentsSpanOutcomeEvaluationEndEvent object`94279428 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.94299430 - `type: "span.outcome_evaluation_end"`94319432 - `id: string`94339434 Unique identifier for this event.94359436 - `explanation: string`94379438 Human-readable explanation of the verdict. For `needs_revision`, describes which criteria failed and why.94399440 - `iteration: number`94419442 0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.94439444 format: int3294459446 - `outcome_evaluation_start_id: string`94479448 The id of the corresponding `span.outcome_evaluation_start` event.94499450 - `outcome_id: string`94519452 The `outc_` ID of the outcome being evaluated.94539454 - `processed_at: string`94559456 A timestamp in RFC 3339 format94579458 format: date-time94599460 - `result: string`94619462 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.94639464 - `usage: BetaManagedAgentsSpanModelUsage`94659466 Token usage for a single model request.94679468 - `cache_creation_input_tokens: number`94699470 Tokens used to create prompt cache in this request.94719472 format: int3294739474 - `cache_read_input_tokens: number`94759476 Tokens read from prompt cache in this request.94779478 format: int3294799480 - `input_tokens: number`94819482 Input tokens consumed by this request.94839484 format: int3294859486 - `output_tokens: number`94879488 Output tokens generated by this request.94899490 format: int3294919492 - `speed: optional "standard" or "fast" or null`94939494 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.94959496 - `"standard"`94979498 - `"fast"`94999500 - `BetaManagedAgentsSpanModelRequestStartEvent object`95019502 Emitted when a model request is initiated by the agent.95039504 - `type: "span.model_request_start"`95059506 - `id: string`95079508 Unique identifier for this event.95099510 - `processed_at: string`95119512 A timestamp in RFC 3339 format95139514 format: date-time95159516 - `BetaManagedAgentsSpanModelRequestEndEvent object`95179518 Emitted when a model request completes.95199520 - `type: "span.model_request_end"`95219522 - `id: string`95239524 Unique identifier for this event.95259526 - `is_error: boolean or null`95279528 Whether the model request resulted in an error.95299530 - `model_request_start_id: string`95319532 The id of the corresponding `span.model_request_start` event.95339534 - `model_usage: BetaManagedAgentsSpanModelUsage`95359536 Token usage for a single model request.95379538 - `processed_at: string`95399540 A timestamp in RFC 3339 format95419542 format: date-time95439544 - `BetaManagedAgentsSpanOutcomeEvaluationOngoingEvent object`95459546 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.95479548 - `type: "span.outcome_evaluation_ongoing"`95499550 - `id: string`95519552 Unique identifier for this event.95539554 - `iteration: number`95559556 0-indexed revision cycle, matching the corresponding `span.outcome_evaluation_start`.95579558 format: int3295599560 - `outcome_id: string`95619562 The `outc_` ID of the outcome being evaluated.95639564 - `processed_at: string`95659566 A timestamp in RFC 3339 format95679568 format: date-time95699570 - `BetaManagedAgentsUserDefineOutcomeEvent object`95719572 Echo of a `user.define_outcome` input event. Carries the server-generated `outcome_id` that subsequent `span.outcome_evaluation_*` events reference.95739574 - `type: "user.define_outcome"`95759576 - `id: string`95779578 Unique identifier for this event.95799580 - `description: string`95819582 What the agent should produce. Copied from the input event.95839584 - `max_iterations: number or null`95859586 Evaluate-then-revise cycles before giving up. Default 3, max 20.95879588 format: int3295899590 - `outcome_id: string`95919592 Server-generated `outc_` ID for this outcome. Referenced by `span.outcome_evaluation_*` events and the session's `outcome_evaluations` list.95939594 - `processed_at: string`95959596 A timestamp in RFC 3339 format95979598 format: date-time95999600 - `rubric: BetaManagedAgentsFileRubric or BetaManagedAgentsTextRubric`96019602 Rubric for grading the quality of an outcome.96039604 - `BetaManagedAgentsFileRubric object`96059606 Rubric referenced by a file uploaded via the Files API.96079608 - `type: "file"`96099610 - `file_id: string`96119612 ID of the rubric file.96139614 - `BetaManagedAgentsTextRubric object`96159616 Rubric content provided inline as text.96179618 - `type: "text"`96199620 - `content: string`96219622 Rubric content. Plain text or markdown — the grader treats it as freeform text.96239624 - `BetaManagedAgentsSessionDeletedEvent object`96259626 Emitted when a session has been deleted. Terminates any active event stream — no further events will be emitted for this session.96279628 - `type: "session.deleted"`96299630 - `id: string`96319632 Unique identifier for this event.96339634 - `processed_at: string`96359636 A timestamp in RFC 3339 format96379638 format: date-time96399640 - `BetaManagedAgentsSessionThreadStatusRunningEvent object`96419642 A session thread has begun executing. Emitted on the thread's own stream and cross-posted to the primary stream for child threads.96439644 - `type: "session.thread_status_running"`96459646 - `id: string`96479648 Unique identifier for this event.96499650 - `agent_name: string`96519652 Name of the agent the thread runs.96539654 - `processed_at: string`96559656 A timestamp in RFC 3339 format96579658 format: date-time96599660 - `session_thread_id: string`96619662 Public sthr_ ID of the thread that started running.96639664 - `BetaManagedAgentsSessionThreadStatusIdleEvent object`96659666 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.96679668 - `type: "session.thread_status_idle"`96699670 - `id: string`96719672 Unique identifier for this event.96739674 - `agent_name: string`96759676 Name of the agent the thread runs.96779678 - `processed_at: string`96799680 A timestamp in RFC 3339 format96819682 format: date-time96839684 - `session_thread_id: string`96859686 Public sthr_ ID of the thread that went idle.96879688 - `stop_reason: BetaManagedAgentsSessionEndTurn or BetaManagedAgentsSessionRequiresAction or BetaManagedAgentsSessionRetriesExhausted or BetaManagedAgentsSessionBudgetReached`96899690 - `BetaManagedAgentsSessionEndTurn object`96919692 The agent completed its turn naturally and is ready for the next user message.96939694 - `BetaManagedAgentsSessionRequiresAction object`96959696 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.96979698 - `BetaManagedAgentsSessionRetriesExhausted object`96999700 The turn ended because repeated errors exhausted the retry budget or an error escalated to `retry_status: 'exhausted'`.97019702 - `BetaManagedAgentsSessionBudgetReached object`97039704 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.97059706 - `BetaManagedAgentsSessionThreadStatusTerminatedEvent object`97079708 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.97099710 - `type: "session.thread_status_terminated"`97119712 - `id: string`97139714 Unique identifier for this event.97159716 - `agent_name: string`97179718 Name of the agent the thread runs.97199720 - `processed_at: string`97219722 A timestamp in RFC 3339 format97239724 format: date-time97259726 - `session_thread_id: string`97279728 Public sthr_ ID of the thread that terminated.97299730 - `BetaManagedAgentsUserToolResultEvent object`97319732 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.97339734 - `type: "user.tool_result"`97359736 - `id: string`97379738 Unique identifier for this event.97399740 - `tool_use_id: string`97419742 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.97439744 - `content: optional array of BetaManagedAgentsTextBlock or BetaManagedAgentsImageBlock or BetaManagedAgentsDocumentBlock or BetaManagedAgentsSearchResultBlock`97459746 The result content returned by the tool.97479748 - `BetaManagedAgentsTextBlock object`97499750 Regular text content.97519752 - `BetaManagedAgentsImageBlock object`97539754 Image content specified directly as base64 data or as a reference via a URL.97559756 - `BetaManagedAgentsDocumentBlock object`97579758 Document content, either specified directly as base64 data, as text, or as a reference via a URL.97599760 - `BetaManagedAgentsSearchResultBlock object`97619762 A block containing a web search result.97639764 - `is_error: optional boolean or null`97659766 Whether the tool execution resulted in an error.97679768 - `processed_at: optional string or null`97699770 A timestamp in RFC 3339 format97719772 format: date-time97739774 - `session_thread_id: optional string or null`97759776 Set by the server to the subagent thread this result was routed to. Omitted when it was routed to the primary thread.97779778 - `BetaManagedAgentsSessionThreadStatusRescheduledEvent object`97799780 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.97819782 - `type: "session.thread_status_rescheduled"`97839784 - `id: string`97859786 Unique identifier for this event.97879788 - `agent_name: string`97899790 Name of the agent the thread runs.97919792 - `processed_at: string`97939794 A timestamp in RFC 3339 format97959796 format: date-time97979798 - `session_thread_id: string`97999800 Public sthr_ ID of the thread that is retrying.98019802 - `BetaManagedAgentsSessionUpdatedEvent object`98039804 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.98059806 - `type: "session.updated"`98079808 - `id: string`98099810 Unique identifier for this event.98119812 - `processed_at: string`98139814 A timestamp in RFC 3339 format98159816 format: date-time98179818 - `agent: optional BetaManagedAgentsSessionAgent or null`98199820 Resolved `agent` definition for a `session`. Snapshot of the `agent` at `session` creation time.98219822 - `type: "agent"`98239824 - `id: string`98259826 - `description: string or null`98279828 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`98299830 - `type: "url"`98319832 - `name: string`98339834 - `url: string`98359836 - `model: BetaManagedAgentsModelConfig`98379838 Model identifier and configuration.98399840 - `id: BetaManagedAgentsModel`98419842 The model that will power your agent.98439844 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.98459846 - `"claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more`98479848 The model that will power your agent.98499850 See [models](https://docs.anthropic.com/en/docs/models-overview) for additional details and options.98519852 - `"claude-fable-5-1"`98539854 Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows98559856 - `"claude-sonnet-5"`98579858 High-performance model for coding and agents98599860 - `"claude-fable-5"`98619862 Next generation of intelligence for the hardest knowledge work and coding problems98639864 - `"claude-opus-5"`98659866 Powerful intelligence for long-running agents and coding98679868 - `"claude-opus-4-8"`98699870 Powerful intelligence for long-running agents and coding98719872 - `"claude-opus-4-7"`98739874 Powerful intelligence for long-running agents and coding98759876 - `"claude-opus-4-6"`98779878 Powerful intelligence for long-running agents and coding98799880 - `"claude-sonnet-4-6"`98819882 Best combination of speed and intelligence98839884 - `"claude-haiku-4-5"`98859886 Fastest model with near-frontier intelligence98879888 - `"claude-haiku-4-5-20251001"`98899890 Fastest model with near-frontier intelligence98919892 - `"claude-opus-4-5"`98939894 Powerful intelligence for long-running agents and coding98959896 - `"claude-opus-4-5-20251101"`98979898 Powerful intelligence for long-running agents and coding98999900 - `"claude-sonnet-4-5"`99019902 High-performance model for agents and coding99039904 - `"claude-sonnet-4-5-20250929"`99059906 High-performance model for agents and coding99079908 - `string`99099910 - `effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more`99119912 How hard Claude works on each turn. Sets `output_config.effort` on every Messages call the session makes.99139914 - `BetaManagedAgentsEffortLow object`99159916 Low effort. Favors latency over reasoning depth.99179918 - `type: "low"`99199920 - `BetaManagedAgentsEffortMedium object`99219922 Medium effort. Balances latency and reasoning depth.99239924 - `type: "medium"`99259926 - `BetaManagedAgentsEffortHigh object`99279928 High effort. Favors reasoning depth.99299930 - `type: "high"`99319932 - `BetaManagedAgentsEffortXhigh object`99339934 Extra-high effort. Not all models accept this level.99359936 - `type: "xhigh"`99379938 - `BetaManagedAgentsEffortMax object`99399940 Maximum effort. Favors reasoning depth over latency.99419942 - `type: "max"`99439944 - `inference_geo: optional string`99459946 Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.99479948 - `speed: optional "standard" or "fast"`99499950 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.99519952 - `"standard"`99539954 - `"fast"`99559956 - `multiagent: BetaManagedAgentsSessionMultiagentCoordinator or null`99579958 Resolved coordinator topology with full agent definitions for each roster member.99599960 - `type: "coordinator"`99619962 - `agents: array of BetaManagedAgentsSessionThreadAgent or BetaManagedAgentsAdvisor`99639964 Full `agent` definitions the coordinator may spawn as session threads.99659966 - `BetaManagedAgentsSessionThreadAgent object`99679968 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`.99699970 - `type: "agent"`99719972 - `id: string`99739974 - `description: string or null`99759976 - `mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition`99779978 - `type: "url"`99799980 - `name: string`99819982 - `url: string`99839984 - `model: BetaManagedAgentsModelConfig`99859986 Model identifier and configuration.99879988 - `name: string`99899990 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`99919992 - `BetaManagedAgentsAnthropicSkill object`99939994 A resolved Anthropic-managed skill.99959996 - `type: "anthropic"`99979998 - `skill_id: string`999910000 - `version: string`1000110002 - `BetaManagedAgentsCustomSkill object`1000310004 A resolved user-created custom skill.1000510006 - `type: "custom"`1000710008 - `skill_id: string`1000910010 - `version: string`1001110012 - `system: string or null`1001310014 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1001510016 - `BetaManagedAgentsAgentToolset20260401 object`1001710018 - `type: "agent_toolset_20260401"`1001910020 - `configs: array of BetaManagedAgentsAgentToolConfig`1002110022 - `BetaManagedAgentsBashToolConfig object`1002310024 Configuration for the bash tool.1002510026 - `type: "bash"`1002710028 - `enabled: boolean`1002910030 - `name: "bash"`1003110032 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1003310034 Permission policy for tool execution.1003510036 - `BetaManagedAgentsAlwaysAllowPolicy object`1003710038 Tool calls are automatically approved without user confirmation.1003910040 - `type: "always_allow"`1004110042 - `BetaManagedAgentsAlwaysAskPolicy object`1004310044 Tool calls require user confirmation before execution.1004510046 - `type: "always_ask"`1004710048 - `BetaManagedAgentsAutoPolicy object`1004910050 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.1005110052 - `type: "auto"`1005310054 - `BetaManagedAgentsEditToolConfig object`1005510056 Configuration for the edit tool.1005710058 - `type: "edit"`1005910060 - `enabled: boolean`1006110062 - `name: "edit"`1006310064 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1006510066 Permission policy for tool execution.1006710068 - `BetaManagedAgentsAlwaysAllowPolicy object`1006910070 Tool calls are automatically approved without user confirmation.1007110072 - `BetaManagedAgentsAlwaysAskPolicy object`1007310074 Tool calls require user confirmation before execution.1007510076 - `BetaManagedAgentsAutoPolicy object`1007710078 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.1007910080 - `BetaManagedAgentsReadToolConfig object`1008110082 Configuration for the read tool.1008310084 - `type: "read"`1008510086 - `enabled: boolean`1008710088 - `name: "read"`1008910090 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1009110092 Permission policy for tool execution.1009310094 - `BetaManagedAgentsAlwaysAllowPolicy object`1009510096 Tool calls are automatically approved without user confirmation.1009710098 - `BetaManagedAgentsAlwaysAskPolicy object`1009910100 Tool calls require user confirmation before execution.1010110102 - `BetaManagedAgentsAutoPolicy object`1010310104 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.1010510106 - `BetaManagedAgentsWriteToolConfig object`1010710108 Configuration for the write tool.1010910110 - `type: "write"`1011110112 - `enabled: boolean`1011310114 - `name: "write"`1011510116 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1011710118 Permission policy for tool execution.1011910120 - `BetaManagedAgentsAlwaysAllowPolicy object`1012110122 Tool calls are automatically approved without user confirmation.1012310124 - `BetaManagedAgentsAlwaysAskPolicy object`1012510126 Tool calls require user confirmation before execution.1012710128 - `BetaManagedAgentsAutoPolicy object`1012910130 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.1013110132 - `BetaManagedAgentsGlobToolConfig object`1013310134 Configuration for the glob tool.1013510136 - `type: "glob"`1013710138 - `enabled: boolean`1013910140 - `name: "glob"`1014110142 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1014310144 Permission policy for tool execution.1014510146 - `BetaManagedAgentsAlwaysAllowPolicy object`1014710148 Tool calls are automatically approved without user confirmation.1014910150 - `BetaManagedAgentsAlwaysAskPolicy object`1015110152 Tool calls require user confirmation before execution.1015310154 - `BetaManagedAgentsAutoPolicy object`1015510156 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.1015710158 - `BetaManagedAgentsGrepToolConfig object`1015910160 Configuration for the grep tool.1016110162 - `type: "grep"`1016310164 - `enabled: boolean`1016510166 - `name: "grep"`1016710168 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1016910170 Permission policy for tool execution.1017110172 - `BetaManagedAgentsAlwaysAllowPolicy object`1017310174 Tool calls are automatically approved without user confirmation.1017510176 - `BetaManagedAgentsAlwaysAskPolicy object`1017710178 Tool calls require user confirmation before execution.1017910180 - `BetaManagedAgentsAutoPolicy object`1018110182 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.1018310184 - `BetaManagedAgentsWebFetchToolConfig object`1018510186 Configuration for the web_fetch tool.1018710188 - `type: "web_fetch"`1018910190 - `enabled: boolean`1019110192 - `name: "web_fetch"`1019310194 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1019510196 Permission policy for tool execution.1019710198 - `BetaManagedAgentsAlwaysAllowPolicy object`1019910200 Tool calls are automatically approved without user confirmation.1020110202 - `BetaManagedAgentsAlwaysAskPolicy object`1020310204 Tool calls require user confirmation before execution.1020510206 - `BetaManagedAgentsAutoPolicy object`1020710208 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.1020910210 - `allowed_domains: optional array of string`1021110212 - `blocked_domains: optional array of string`1021310214 - `max_content_tokens: optional number or null`1021510216 format: int321021710218 - `BetaManagedAgentsWebSearchToolConfig object`1021910220 Configuration for the web_search tool.1022110222 - `type: "web_search"`1022310224 - `enabled: boolean`1022510226 - `name: "web_search"`1022710228 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1022910230 Permission policy for tool execution.1023110232 - `BetaManagedAgentsAlwaysAllowPolicy object`1023310234 Tool calls are automatically approved without user confirmation.1023510236 - `BetaManagedAgentsAlwaysAskPolicy object`1023710238 Tool calls require user confirmation before execution.1023910240 - `BetaManagedAgentsAutoPolicy object`1024110242 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.1024310244 - `allowed_domains: optional array of string`1024510246 - `blocked_domains: optional array of string`1024710248 - `user_location: optional BetaManagedAgentsUserLocation or null`1024910250 Approximate user location for search result localization.1025110252 - `type: "approximate"`1025310254 Location precision. Only "approximate" is supported.1025510256 - `city: optional string or null`1025710258 City name.1025910260 minLength: 1, maxLength: 2551026110262 - `country: optional string or null`1026310264 Two-letter ISO 3166-1 country code, uppercase.1026510266 - `region: optional string or null`1026710268 Region or state name.1026910270 minLength: 1, maxLength: 2551027110272 - `timezone: optional string or null`1027310274 IANA timezone identifier, e.g. "America/Los_Angeles".1027510276 minLength: 1, maxLength: 2551027710278 - `default_config: BetaManagedAgentsAgentToolsetDefaultConfig`1027910280 Resolved default configuration for agent tools.1028110282 - `enabled: boolean`1028310284 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1028510286 Permission policy for tool execution.1028710288 - `BetaManagedAgentsAlwaysAllowPolicy object`1028910290 Tool calls are automatically approved without user confirmation.1029110292 - `BetaManagedAgentsAlwaysAskPolicy object`1029310294 Tool calls require user confirmation before execution.1029510296 - `BetaManagedAgentsAutoPolicy object`1029710298 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.1029910300 - `BetaManagedAgentsMCPToolset object`1030110302 - `type: "mcp_toolset"`1030310304 - `configs: array of BetaManagedAgentsMCPToolConfig`1030510306 - `enabled: boolean`1030710308 - `name: string`1030910310 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1031110312 Permission policy for tool execution.1031310314 - `BetaManagedAgentsAlwaysAllowPolicy object`1031510316 Tool calls are automatically approved without user confirmation.1031710318 - `BetaManagedAgentsAlwaysAskPolicy object`1031910320 Tool calls require user confirmation before execution.1032110322 - `BetaManagedAgentsAutoPolicy object`1032310324 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.1032510326 - `default_config: BetaManagedAgentsMCPToolsetDefaultConfig`1032710328 Resolved default configuration for all tools from an MCP server.1032910330 - `enabled: boolean`1033110332 - `permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy`1033310334 Permission policy for tool execution.1033510336 - `BetaManagedAgentsAlwaysAllowPolicy object`1033710338 Tool calls are automatically approved without user confirmation.1033910340 - `BetaManagedAgentsAlwaysAskPolicy object`1034110342 Tool calls require user confirmation before execution.1034310344 - `BetaManagedAgentsAutoPolicy object`1034510346 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.1034710348 - `mcp_server_name: string`1034910350 - `BetaManagedAgentsCustomTool object`1035110352 A custom tool as returned in API responses.1035310354 - `type: "custom"`1035510356 - `description: string`1035710358 - `input_schema: BetaManagedAgentsCustomToolInputSchema`1035910360 JSON Schema for custom tool input parameters.1036110362 - `type: "object"`1036310364 - `properties: optional map[unknown] or null`1036510366 - `required: optional array of string or null`1036710368 - `name: string`1036910370 - `version: number`1037110372 format: int321037310374 - `BetaManagedAgentsAdvisor object`1037510376 Platform advisor roster entry: a model the session's primary thread may consult mid-turn.1037710378 - `type: "advisor"`1037910380 - `model: string`1038110382 The advisor model id.1038310384 - `name: string`1038510386 - `skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill`1038710388 - `BetaManagedAgentsAnthropicSkill object`1038910390 A resolved Anthropic-managed skill.1039110392 - `BetaManagedAgentsCustomSkill object`1039310394 A resolved user-created custom skill.1039510396 - `system: string or null`1039710398 - `tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool`1039910400 - `BetaManagedAgentsAgentToolset20260401 object`1040110402 - `BetaManagedAgentsMCPToolset object`1040310404 - `BetaManagedAgentsCustomTool object`1040510406 A custom tool as returned in API responses.1040710408 - `version: number`1040910410 format: int321041110412 - `budget: optional BetaManagedAgentsBudgetLimit or null`1041310414 A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1041510416 - `type: "limit"`1041710418 - `max_list_cost: BetaMonetaryAmount`1041910420 A monetary amount in a specific currency.1042110422 - `amount: string`1042310424 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.1042510426 - `currency: BetaCurrency`1042710428 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.1042910430 - `metadata: optional map[string]`1043110432 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.1043310434 - `title: optional string or null`1043510436 The session's new title. Present only when the update changed it.1043710438 - `BetaManagedAgentsStartEvent object`1043910440 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.1044110442 - `type: "event_start"`1044310444 - `event: BetaManagedAgentsStartEventPreview`1044510446 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.1044710448 - `BetaManagedAgentsAgentMessagePreview object`1044910450 - `type: "agent.message"`1045110452 - `id: string`1045310454 The id the buffered agent.message will carry if it is emitted. Matches the event_id on this preview's event_delta events.1045510456 - `BetaManagedAgentsAgentThinkingPreview object`1045710458 - `type: "agent.thinking"`1045910460 - `id: string`1046110462 The id the buffered agent.thinking will carry if it is emitted. Start-only — no event_delta events follow.1046310464 - `BetaManagedAgentsDeltaEvent object`1046510466 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.1046710468 - `type: "event_delta"`1046910470 - `delta: BetaManagedAgentsDeltaContent`1047110472 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.1047310474 - `type: "content_delta"`1047510476 - `content: BetaManagedAgentsTextBlock`1047710478 Regular text content.1047910480 - `index: optional number`1048110482 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.1048310484 format: uint321048510486 - `event_id: string`1048710488 The id of the event being previewed. Matches event.id on the corresponding event_start and the buffered event that reconciles the preview.1048910490 - `BetaManagedAgentsSystemMessageEvent object`1049110492 A mid-conversation system message event. Carries system-role content that is appended to the session as a `role: "system"` turn.1049310494 - `type: "system.message"`1049510496 - `id: string`1049710498 Unique identifier for this event.1049910500 - `content: array of BetaManagedAgentsSystemContentBlock`1050110502 System content blocks. Text-only.1050310504 - `type: "text"`1050510506 - `text: string`1050710508 The text content.1050910510 minLength: 11051110512 - `processed_at: optional string or null`1051310514 A timestamp in RFC 3339 format1051510516 format: date-time1051710518 - `BetaManagedAgentsSessionUsageEvent object`1051910520 Periodic snapshot of the session's cumulative usage and tracked list cost.1052110522 - `type: "session.usage"`1052310524 - `id: string`1052510526 Unique identifier for this event.1052710528 - `processed_at: string`1052910530 A timestamp in RFC 3339 format1053110532 format: date-time1053310534 - `usage: BetaManagedAgentsSessionUsageSnapshot`1053510536 Point-in-time snapshot of a session's cumulative usage.1053710538 - `active_seconds: optional number`1053910540 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.1054110542 format: double1054310544 - `cache_creation: optional BetaManagedAgentsCacheCreationUsage`1054510546 Prompt-cache creation token usage broken down by cache lifetime.1054710548 - `ephemeral_1h_input_tokens: optional number`1054910550 Tokens used to create 1-hour ephemeral cache entries.1055110552 format: int321055310554 - `ephemeral_5m_input_tokens: optional number`1055510556 Tokens used to create 5-minute ephemeral cache entries.1055710558 format: int321055910560 - `cache_read_input_tokens: optional number`1056110562 Total tokens read from prompt cache.1056310564 format: int321056510566 - `input_tokens: optional number`1056710568 Total input tokens consumed across all turns.1056910570 format: int321057110572 - `list_cost: optional BetaMonetaryAmount`1057310574 A monetary amount in a specific currency.1057510576 - `output_tokens: optional number`1057710578 Total output tokens generated across all turns.1057910580 format: int321058110582 - `server_tool_use: optional BetaManagedAgentsServerToolUsage`1058310584 Cumulative count of server-executed tool invocations, broken down by tool.1058510586 - `web_fetch_requests: optional number`1058710588 Number of server-executed web fetch requests.1058910590 format: int321059110592 - `web_search_requests: optional number`1059310594 Number of server-executed web search requests.1059510596 format: int321059710598 - `budget: optional BetaManagedAgentsBudgetLimit or null`1059910600 A hard spend ceiling. The session stops issuing new model requests once the tracked list cost reaches `max_list_cost`.1060110602#### Example1060310604```bash10605curl https://api.anthropic.com/v1/sessions/$SESSION_ID/threads/$THREAD_ID/stream \10606 -H 'anthropic-version: 2023-06-01' \10607 -H 'anthropic-beta: managed-agents-2026-04-01' \10608 -H "X-Api-Key: $ANTHROPIC_API_KEY"10609```1061010611##### Response (200)1061210613```json10614{10615 "id": "sevt_011CZkZGOp0iBcp4kaQSihUmy",10616 "content": [10617 {10618 "text": "Where is my order #1234?",10619 "type": "text"10620 }10621 ],10622 "type": "user.message",10623 "processed_at": "2026-03-15T10:00:00Z"10624}10625```10626