SPB Git
spboucher.ai

Repository Public

spbdoc-api

create.md 47.0 KB

# title: Create Agent url: https://platform.claude.com/docs/en/api/beta/agents/create

# Create Agent

POST /v1/agents

Create Agent

# Headers

  • "anthropic-beta": optional array of AnthropicBeta

    Optional header to specify the beta version(s) you want to use.

    • string

    • "message-batches-2024-09-24" or "prompt-caching-2024-07-31" or "computer-use-2024-10-22" or 43 more

      • "message-batches-2024-09-24"

      • "prompt-caching-2024-07-31"

      • "computer-use-2024-10-22"

      • "computer-use-2025-01-24"

      • "pdfs-2024-09-25"

      • "token-counting-2024-11-01"

      • "token-efficient-tools-2025-02-19"

      • "output-128k-2025-02-19"

      • "files-api-2025-04-14"

      • "mcp-client-2025-04-04"

      • "mcp-client-2025-11-20"

      • "dev-full-thinking-2025-05-14"

      • "interleaved-thinking-2025-05-14"

      • "code-execution-2025-05-22"

      • "extended-cache-ttl-2025-04-11"

      • "context-1m-2025-08-07"

      • "context-management-2025-06-27"

      • "model-context-window-exceeded-2025-08-26"

      • "skills-2025-10-02"

      • "fast-mode-2026-02-01"

      • "output-300k-2026-03-24"

      • "user-profiles-2026-03-24"

      • "user-profiles-2026-08-18"

      • "user-profiles-2026-09-04"

      • "advisor-tool-2026-03-01"

      • "managed-agents-2026-04-01"

      • "cache-diagnosis-2026-04-07"

      • "dreaming-2026-04-21"

      • "thinking-token-count-2026-05-13"

      • "server-side-fallback-2026-06-01"

      • "server-side-fallback-2026-07-01"

      • "fallback-credit-2026-06-01"

      • "fallback-credit-2026-07-01"

      • "agent-memory-2026-07-22"

      • "mid-conversation-tool-changes-2026-07-01"

      • "compact-2026-01-12"

      • "computer-use-2025-11-24"

      • "mcp-tunnels-2026-06-22"

      • "structured-outputs-2025-11-13"

      • "task-budgets-2026-03-13"

      • "thinking-display-updates-2026-08-18"

      • "ce-user-management-2026-07-13"

      • "mid-conversation-output-config-2026-07-01"

      • "thinking-binding-controls-2026-08-01"

      • "mid-conversation-system-clear-at-2026-08-21"

      • "compact-2026-09-04"

  • "anthropic-workspace-id": optional string

# Body parameters

  • model: BetaManagedAgentsModel or BetaManagedAgentsModelConfigParams

    Model identifier. Accepts the model string, e.g. claude-opus-5, or a model_config object for additional configuration control

    • BetaManagedAgentsModel = "claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more or string

      The model that will power your agent.

      See models for additional details and options.

      • "claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more

        The model that will power your agent.

        See models for additional details and options.

        • "claude-fable-5-1"

          Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows

        • "claude-sonnet-5"

          High-performance model for coding and agents

        • "claude-fable-5"

          Next generation of intelligence for the hardest knowledge work and coding problems

        • "claude-opus-5"

          Powerful intelligence for long-running agents and coding

        • "claude-opus-4-8"

          Powerful intelligence for long-running agents and coding

        • "claude-opus-4-7"

          Powerful intelligence for long-running agents and coding

        • "claude-opus-4-6"

          Powerful intelligence for long-running agents and coding

        • "claude-sonnet-4-6"

          Best combination of speed and intelligence

        • "claude-haiku-4-5"

          Fastest model with near-frontier intelligence

        • "claude-haiku-4-5-20251001"

          Fastest model with near-frontier intelligence

        • "claude-opus-4-5"

          Powerful intelligence for long-running agents and coding

        • "claude-opus-4-5-20251101"

          Powerful intelligence for long-running agents and coding

        • "claude-sonnet-4-5"

          High-performance model for agents and coding

        • "claude-sonnet-4-5-20250929"

          High-performance model for agents and coding

      • string

    • BetaManagedAgentsModelConfigParams object

      An object that defines additional configuration control over model use

      • id: BetaManagedAgentsModel

        The model that will power your agent.

        See models for additional details and options.

      • effort: optional "low" or "medium" or "high" or 2 more or BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or 3 more or null

        How hard Claude works on each inference call. Accepts a bare level string ("high") or {"type": "high"}. On create, omitting it resolves the per-model default; on update, omitting it leaves the stored value unchanged.

        • BetaManagedAgentsEffortLevel = "low" or "medium" or "high" or 2 more

          How hard Claude works on each turn. Higher levels favor reasoning depth over latency. Not all models accept every level; invalid combinations are rejected at create time.

          • "low"

          • "medium"

          • "high"

          • "xhigh"

          • "max"

        • BetaManagedAgentsEffortLow object

          Low effort. Favors latency over reasoning depth.

          • type: "low"

        • BetaManagedAgentsEffortMedium object

          Medium effort. Balances latency and reasoning depth.

          • type: "medium"

        • BetaManagedAgentsEffortHigh object

          High effort. Favors reasoning depth.

          • type: "high"

        • BetaManagedAgentsEffortXhigh object

          Extra-high effort. Not all models accept this level.

          • type: "xhigh"

        • BetaManagedAgentsEffortMax object

          Maximum effort. Favors reasoning depth over latency.

          • type: "max"

      • inference_geo: optional string or null

        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo. On update, model is whole-object replacement — omitting inference_geo clears it.

      • speed: optional "standard" or "fast" or null

        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.

        • "standard"

        • "fast"

  • name: string

    Human-readable name for the agent.

    minLength: 1, maxLength: 256

  • description: optional string or null

    Description of what the agent does.

    maxLength: 2048

  • mcp_servers: optional array of BetaManagedAgentsURLMCPServerParams

    MCP servers this agent connects to. Maximum 20. Names must be unique within the array. Every server must be referenced by an mcp_toolset in tools; unreferenced servers are rejected. See the MCP connector guide.

    • type: "url"

    • name: string

      Unique name for this server, referenced by mcp_toolset configurations. 1-255 characters.

      minLength: 1, maxLength: 255

    • url: string

      Endpoint URL for the MCP server.

      maxLength: 2048

  • metadata: optional map[string]

    Arbitrary key-value metadata. Maximum 16 pairs, keys up to 64 chars, values up to 512 chars.

  • multiagent: optional BetaManagedAgentsMultiagentParams or null

    A coordinator topology: the session's primary thread orchestrates work by spawning session threads, each running an agent drawn from the agents roster.

    • type: "coordinator"

    • agents: array of BetaManagedAgentsMultiagentRosterEntryParams

      Agents the coordinator may spawn as session threads. 1–20 entries. Each entry is an agent ID string, a versioned {"type":"agent","id","version"} reference, or {"type":"self"} to allow recursive self-invocation. Entries must reference distinct agents (after resolving self and string forms); at most one self. Referenced agents must exist, must not be archived, and must not themselves have multiagent set (depth limit 1).

      • string

      • BetaManagedAgentsAgentParams object

        Specification for an Agent. Provide a specific version or use the short-form agent="agent_id" for the most recent version

        • type: "agent"

        • id: string

          The agent ID.

          minLength: 1, maxLength: 128

        • version: optional number

          The specific agent version to use. Omit to use the latest version. Must be at least 1 if specified.

          format: int32

      • BetaManagedAgentsMultiagentSelfParams object

        Sentinel roster entry meaning "the agent that owns this configuration". Resolved server-side to a concrete agent reference.

        • type: "self"

      • BetaManagedAgentsAdvisorParams object

        Platform advisor roster entry: a model the session's primary thread may consult mid-turn. At most one per roster; the entry occupies the roster name anthropic.advisor.

        • type: "advisor"

        • model: string

          A Claude model id. The model must be permitted as an advisor for this agent's model — see the sessions/threads/advisor spec.

          minLength: 1, maxLength: 256

  • skills: optional array of BetaManagedAgentsSkillParams

    Skills available to the agent.

    • BetaManagedAgentsAnthropicSkillParams object

      An Anthropic-managed skill.

      • type: "anthropic"

      • skill_id: string

        Identifier of the Anthropic skill (e.g., "xlsx").

        minLength: 1, maxLength: 64

      • version: optional string or null

        Version to pin. Defaults to latest if omitted.

        minLength: 1, maxLength: 64

    • BetaManagedAgentsCustomSkillParams object

      A user-created custom skill.

      • type: "custom"

      • skill_id: string

        Tagged ID of the custom skill (e.g., "skill_01XJ5...").

        minLength: 1, maxLength: 64

      • version: optional string or null

        Version to pin. Defaults to latest if omitted.

        minLength: 1, maxLength: 64

  • system: optional string or null

    System prompt for the agent.

    maxLength: 100000

  • tools: optional array of BetaManagedAgentsAgentToolset20260401Params or BetaManagedAgentsMCPToolsetParams or BetaManagedAgentsCustomToolParams

    Tool configurations available to the agent. Maximum of 128 tools across all toolsets allowed.

    • BetaManagedAgentsAgentToolset20260401Params object

      Configuration for built-in agent tools. Use this to enable or disable groups of tools available to the agent.

      • type: "agent_toolset_20260401"

      • configs: optional array of BetaManagedAgentsAgentToolConfigParams

        Per-tool configuration overrides.

        • BetaManagedAgentsBashToolConfigParams object

          Configuration override for the bash tool.

          • type: optional "bash"

          • name: "bash"

            Must be "bash".

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

              • type: "always_allow"

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

              • type: "always_ask"

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

              • type: "auto"

        • BetaManagedAgentsEditToolConfigParams object

          Configuration override for the edit tool.

          • type: optional "edit"

          • name: "edit"

            Must be "edit".

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • BetaManagedAgentsReadToolConfigParams object

          Configuration override for the read tool.

          • type: optional "read"

          • name: "read"

            Must be "read".

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • BetaManagedAgentsWriteToolConfigParams object

          Configuration override for the write tool.

          • type: optional "write"

          • name: "write"

            Must be "write".

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • BetaManagedAgentsGlobToolConfigParams object

          Configuration override for the glob tool.

          • type: optional "glob"

          • name: "glob"

            Must be "glob".

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • BetaManagedAgentsGrepToolConfigParams object

          Configuration override for the grep tool.

          • type: optional "grep"

          • name: "grep"

            Must be "grep".

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • BetaManagedAgentsWebFetchToolConfigParams object

          Configuration override for the web_fetch tool.

          • type: optional "web_fetch"

          • name: "web_fetch"

            Must be "web_fetch".

          • allowed_domains: optional array of string

            Only fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.

          • blocked_domains: optional array of string

            Never fetch URLs whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme, port, or path). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • max_content_tokens: optional number or null

            Maximum number of tokens of fetched text content to include in context per call. Does not apply to binary content such as PDFs.

            format: int32

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • BetaManagedAgentsWebSearchToolConfigParams object

          Configuration override for the web_search tool.

          • type: optional "web_search"

          • name: "web_search"

            Must be "web_search".

          • allowed_domains: optional array of string

            Only return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "docs.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with blocked_domains.

          • blocked_domains: optional array of string

            Never return search results whose host is one of these domains or a subdomain of one. Each entry is a plain hostname like "ads.example.com" (no scheme or port; an optional path suffix is accepted). At most 64 entries; an empty list is rejected (omit the field instead). Cannot be combined with allowed_domains.

          • enabled: optional boolean or null

            Whether this tool is enabled and available to Claude. Overrides the default_config setting.

          • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

          • user_location: optional BetaManagedAgentsUserLocation or null

            Approximate user location for search result localization.

            • type: "approximate"

              Location precision. Only "approximate" is supported.

            • city: optional string or null

              City name.

              minLength: 1, maxLength: 255

            • country: optional string or null

              Two-letter ISO 3166-1 country code, uppercase.

            • region: optional string or null

              Region or state name.

              minLength: 1, maxLength: 255

            • timezone: optional string or null

              IANA timezone identifier, e.g. "America/Los_Angeles".

              minLength: 1, maxLength: 255

      • default_config: optional BetaManagedAgentsAgentToolsetDefaultConfigParams or null

        Default configuration for all tools in a toolset.

        • enabled: optional boolean or null

          Whether tools are enabled and available to Claude by default. Defaults to true if not specified.

        • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

          Permission policy for tool execution.

          • BetaManagedAgentsAlwaysAllowPolicy object

            Tool calls are automatically approved without user confirmation.

          • BetaManagedAgentsAlwaysAskPolicy object

            Tool calls require user confirmation before execution.

          • BetaManagedAgentsAutoPolicy object

            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

    • BetaManagedAgentsMCPToolsetParams object

      Configuration for tools from an MCP server defined in mcp_servers.

      • type: "mcp_toolset"

      • mcp_server_name: string

        Name of the MCP server. Must match a server name from the mcp_servers array. 1-255 characters.

        minLength: 1, maxLength: 255

      • configs: optional array of BetaManagedAgentsMCPToolConfigParams

        Per-tool configuration overrides.

        • name: string

          Name of the MCP tool to configure. 1-128 characters.

          minLength: 1, maxLength: 128

        • enabled: optional boolean or null

          Whether this tool is enabled. Overrides the default_config setting.

        • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

          Permission policy for tool execution.

          • BetaManagedAgentsAlwaysAllowPolicy object

            Tool calls are automatically approved without user confirmation.

          • BetaManagedAgentsAlwaysAskPolicy object

            Tool calls require user confirmation before execution.

          • BetaManagedAgentsAutoPolicy object

            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

      • default_config: optional BetaManagedAgentsMCPToolsetDefaultConfigParams or null

        Default configuration for all tools from an MCP server.

        • enabled: optional boolean or null

          Whether tools are enabled by default. Defaults to true if not specified.

        • permission_policy: optional BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy or null

          Permission policy for tool execution.

          • BetaManagedAgentsAlwaysAllowPolicy object

            Tool calls are automatically approved without user confirmation.

          • BetaManagedAgentsAlwaysAskPolicy object

            Tool calls require user confirmation before execution.

          • BetaManagedAgentsAutoPolicy object

            The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

    • BetaManagedAgentsCustomToolParams object

      A custom tool that is executed by the API client rather than the agent. When the agent calls this tool, an agent.custom_tool_use event is emitted and the session goes idle, waiting for the client to provide the result via a user.custom_tool_result event.

      • type: "custom"

      • description: string

        Description of what the tool does, shown to the agent to help it decide when to use the tool.

        minLength: 1

      • input_schema: BetaManagedAgentsCustomToolInputSchema

        JSON Schema for custom tool input parameters.

        • type: "object"

        • properties: optional map[unknown] or null

        • required: optional array of string or null

      • name: string

        Unique name for the tool. 1-128 characters; letters, digits, underscores, and hyphens.

        minLength: 1, maxLength: 128

# Returns

  • BetaManagedAgentsAgent object

    A Managed Agents agent.

    • type: "agent"

    • id: string

    • archived_at: string or null

      A timestamp in RFC 3339 format

      format: date-time

    • created_at: string

      A timestamp in RFC 3339 format

      format: date-time

    • description: string or null

    • mcp_servers: array of BetaManagedAgentsMCPServerURLDefinition

      • type: "url"

      • name: string

      • url: string

    • metadata: map[string]

    • model: BetaManagedAgentsModelConfig

      Model identifier and configuration.

      • id: BetaManagedAgentsModel

        The model that will power your agent.

        See models for additional details and options.

        • "claude-fable-5-1" or "claude-sonnet-5" or "claude-fable-5" or 11 more

          The model that will power your agent.

          See models for additional details and options.

          • "claude-fable-5-1"

            Frontier intelligence for ambitious tasks across coding, scientific discovery, and enterprise workflows

          • "claude-sonnet-5"

            High-performance model for coding and agents

          • "claude-fable-5"

            Next generation of intelligence for the hardest knowledge work and coding problems

          • "claude-opus-5"

            Powerful intelligence for long-running agents and coding

          • "claude-opus-4-8"

            Powerful intelligence for long-running agents and coding

          • "claude-opus-4-7"

            Powerful intelligence for long-running agents and coding

          • "claude-opus-4-6"

            Powerful intelligence for long-running agents and coding

          • "claude-sonnet-4-6"

            Best combination of speed and intelligence

          • "claude-haiku-4-5"

            Fastest model with near-frontier intelligence

          • "claude-haiku-4-5-20251001"

            Fastest model with near-frontier intelligence

          • "claude-opus-4-5"

            Powerful intelligence for long-running agents and coding

          • "claude-opus-4-5-20251101"

            Powerful intelligence for long-running agents and coding

          • "claude-sonnet-4-5"

            High-performance model for agents and coding

          • "claude-sonnet-4-5-20250929"

            High-performance model for agents and coding

        • string

      • effort: optional BetaManagedAgentsEffortLow or BetaManagedAgentsEffortMedium or BetaManagedAgentsEffortHigh or 2 more

        How hard Claude works on each turn. Sets output_config.effort on every Messages call the session makes.

        • BetaManagedAgentsEffortLow object

          Low effort. Favors latency over reasoning depth.

          • type: "low"

        • BetaManagedAgentsEffortMedium object

          Medium effort. Balances latency and reasoning depth.

          • type: "medium"

        • BetaManagedAgentsEffortHigh object

          High effort. Favors reasoning depth.

          • type: "high"

        • BetaManagedAgentsEffortXhigh object

          Extra-high effort. Not all models accept this level.

          • type: "xhigh"

        • BetaManagedAgentsEffortMax object

          Maximum effort. Favors reasoning depth over latency.

          • type: "max"

      • inference_geo: optional string

        Geographic region for model inference. When unset, requests fall through to the workspace's default_inference_geo.

      • speed: optional "standard" or "fast"

        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.

        • "standard"

        • "fast"

    • multiagent: BetaManagedAgentsMultiagent or null

      Resolved coordinator topology with a concrete agent roster.

      • type: "coordinator"

      • agents: array of BetaManagedAgentsAgentReference or BetaManagedAgentsAdvisor

        Agents the coordinator may spawn as session threads, each resolved to a specific version.

        • BetaManagedAgentsAgentReference object

          A resolved agent reference with a concrete version.

          • type: "agent"

          • id: string

          • version: number

            format: int32

        • BetaManagedAgentsAdvisor object

          Platform advisor roster entry: a model the session's primary thread may consult mid-turn.

          • type: "advisor"

          • model: string

            The advisor model id.

    • name: string

    • skills: array of BetaManagedAgentsAnthropicSkill or BetaManagedAgentsCustomSkill

      • BetaManagedAgentsAnthropicSkill object

        A resolved Anthropic-managed skill.

        • type: "anthropic"

        • skill_id: string

        • version: string

      • BetaManagedAgentsCustomSkill object

        A resolved user-created custom skill.

        • type: "custom"

        • skill_id: string

        • version: string

    • system: string or null

    • tools: array of BetaManagedAgentsAgentToolset20260401 or BetaManagedAgentsMCPToolset or BetaManagedAgentsCustomTool

      • BetaManagedAgentsAgentToolset20260401 object

        • type: "agent_toolset_20260401"

        • configs: array of BetaManagedAgentsAgentToolConfig

          • BetaManagedAgentsBashToolConfig object

            Configuration for the bash tool.

            • type: "bash"

            • enabled: boolean

            • name: "bash"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

                • type: "always_allow"

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

                • type: "always_ask"

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

                • type: "auto"

          • BetaManagedAgentsEditToolConfig object

            Configuration for the edit tool.

            • type: "edit"

            • enabled: boolean

            • name: "edit"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

          • BetaManagedAgentsReadToolConfig object

            Configuration for the read tool.

            • type: "read"

            • enabled: boolean

            • name: "read"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

          • BetaManagedAgentsWriteToolConfig object

            Configuration for the write tool.

            • type: "write"

            • enabled: boolean

            • name: "write"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

          • BetaManagedAgentsGlobToolConfig object

            Configuration for the glob tool.

            • type: "glob"

            • enabled: boolean

            • name: "glob"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

          • BetaManagedAgentsGrepToolConfig object

            Configuration for the grep tool.

            • type: "grep"

            • enabled: boolean

            • name: "grep"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

          • BetaManagedAgentsWebFetchToolConfig object

            Configuration for the web_fetch tool.

            • type: "web_fetch"

            • enabled: boolean

            • name: "web_fetch"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

            • allowed_domains: optional array of string

            • blocked_domains: optional array of string

            • max_content_tokens: optional number or null

              format: int32

          • BetaManagedAgentsWebSearchToolConfig object

            Configuration for the web_search tool.

            • type: "web_search"

            • enabled: boolean

            • name: "web_search"

            • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

              Permission policy for tool execution.

              • BetaManagedAgentsAlwaysAllowPolicy object

                Tool calls are automatically approved without user confirmation.

              • BetaManagedAgentsAlwaysAskPolicy object

                Tool calls require user confirmation before execution.

              • BetaManagedAgentsAutoPolicy object

                The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

            • allowed_domains: optional array of string

            • blocked_domains: optional array of string

            • user_location: optional BetaManagedAgentsUserLocation or null

              Approximate user location for search result localization.

              • type: "approximate"

                Location precision. Only "approximate" is supported.

              • city: optional string or null

                City name.

                minLength: 1, maxLength: 255

              • country: optional string or null

                Two-letter ISO 3166-1 country code, uppercase.

              • region: optional string or null

                Region or state name.

                minLength: 1, maxLength: 255

              • timezone: optional string or null

                IANA timezone identifier, e.g. "America/Los_Angeles".

                minLength: 1, maxLength: 255

        • default_config: BetaManagedAgentsAgentToolsetDefaultConfig

          Resolved default configuration for agent tools.

          • enabled: boolean

          • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

      • BetaManagedAgentsMCPToolset object

        • type: "mcp_toolset"

        • configs: array of BetaManagedAgentsMCPToolConfig

          • enabled: boolean

          • name: string

          • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • default_config: BetaManagedAgentsMCPToolsetDefaultConfig

          Resolved default configuration for all tools from an MCP server.

          • enabled: boolean

          • permission_policy: BetaManagedAgentsAlwaysAllowPolicy or BetaManagedAgentsAlwaysAskPolicy or BetaManagedAgentsAutoPolicy

            Permission policy for tool execution.

            • BetaManagedAgentsAlwaysAllowPolicy object

              Tool calls are automatically approved without user confirmation.

            • BetaManagedAgentsAlwaysAskPolicy object

              Tool calls require user confirmation before execution.

            • BetaManagedAgentsAutoPolicy object

              The server decides each tool call individually: it judges, from the tool, its input, and the session content so far, whether the call is safe to execute or high-risk, and evaluates it to allow when judged safe and to deny when judged high-risk. A call the server cannot reach a judgement on evaluates to ask.

        • mcp_server_name: string

      • BetaManagedAgentsCustomTool object

        A custom tool as returned in API responses.

        • type: "custom"

        • description: string

        • input_schema: BetaManagedAgentsCustomToolInputSchema

          JSON Schema for custom tool input parameters.

          • type: "object"

          • properties: optional map[unknown] or null

          • required: optional array of string or null

        • name: string

    • updated_at: string

      A timestamp in RFC 3339 format

      format: date-time

    • version: number

      The agent's current version. Starts at 1 and increments when the agent is modified.

      format: int32

# Example

bash
curl https://api.anthropic.com/v1/agents \
    -H 'Content-Type: application/json' \
    -H 'anthropic-version: 2023-06-01' \
    -H 'anthropic-beta: managed-agents-2026-04-01' \
    -H "X-Api-Key: $ANTHROPIC_API_KEY" \
    -d '{
          "model": "claude-opus-5",
          "name": "My First Agent",
          "description": "A general-purpose starter agent.",
          "metadata": {
            "foo": "bar"
          },
          "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user'\''s task end to end.",
          "tools": [
            {
              "type": "agent_toolset_20260401"
            }
          ]
        }'

# Response (200)

json
{
  "id": "agent_011CZkYpogX7uDKUyvBTophP",
  "archived_at": null,
  "created_at": "2026-03-15T10:00:00Z",
  "description": "A general-purpose starter agent.",
  "mcp_servers": [
    {
      "name": "example-mcp",
      "type": "url",
      "url": "https://example-server.modelcontextprotocol.io/sse"
    }
  ],
  "metadata": {
    "foo": "bar"
  },
  "model": {
    "id": "claude-opus-5",
    "effort": {
      "type": "low"
    },
    "inference_geo": "inference_geo",
    "speed": "standard"
  },
  "multiagent": {
    "agents": [
      {
        "id": "agent_011CZkYqphY8vELVzwCUpqiQ",
        "type": "agent",
        "version": 1
      }
    ],
    "type": "coordinator"
  },
  "name": "My First Agent",
  "skills": [
    {
      "skill_id": "xlsx",
      "type": "anthropic",
      "version": "1"
    },
    {
      "skill_id": "skill_011CZkZFNu9hAbo3jZPRgTlx",
      "type": "custom",
      "version": "2"
    }
  ],
  "system": "You are a general-purpose agent that can research, write code, run commands, and use connected tools to complete the user's task end to end.",
  "tools": [
    {
      "configs": [
        {
          "enabled": true,
          "name": "bash",
          "permission_policy": {
            "type": "always_allow"
          },
          "type": "bash"
        }
      ],
      "default_config": {
        "enabled": true,
        "permission_policy": {
          "type": "always_ask"
        }
      },
      "type": "agent_toolset_20260401"
    }
  ],
  "type": "agent",
  "updated_at": "2026-03-15T10:00:00Z",
  "version": 1
}
Files main
Menu

Theme
↑↓navigate ↵open escclose /⌘Kanywhere