Body

• access_control object

  Additional properties are NOT allowed.

  Hide access_control attribute Show access_control attribute object

  • access_mode string Required Generally available

    Access-control mode: public (any privileged user can read/write), shared (any privileged user can read, only owner can write), private (only owner can read/write). Agents created without an access-control mode default to private.

    Values are public, shared, or private.

• avatar_color string

  Optional hex color code for the agent avatar.

• avatar_symbol string

  Optional symbol/initials for the agent avatar.

• configuration object Required

  Configuration settings for the agent.

  Additional properties are NOT allowed.

  Hide configuration attributes Show configuration attributes object

  • ai_indices array[string]

    Array of AI indices to associate with the agent.

    Not more than 100 elements. Maximum length of each is 256.

  • approvals object Generally available

    Auto-approval defaults applied to every run of this agent.

    Additional properties are NOT allowed.

    Hide approvals attribute Show approvals attribute object

    • auto_approved_apis object

      Destructive Elasticsearch or Kibana APIs this agent may call without a user confirmation, keyed by backend.

      Additional properties are NOT allowed.

      Hide auto_approved_apis attributes Show auto_approved_apis attributes object

      • elasticsearch array[string]

        Elasticsearch APIs pre-approved for every run of this agent. Each entry is an exact identifier formed from the namespace and name (for example indices.create), a namespace wildcard (for example indices.*), or * for every Elasticsearch API. Entries are not checked against the known APIs, and one that matches no API has no effect.

        Not more than 100 elements. Maximum length of each is 256.

      • kibana array[string]

        Kibana APIs pre-approved for every run of this agent. Each entry is an exact identifier formed from the namespace and name (for example alerting.delete-alerting-rule-id), a namespace wildcard (for example alerting.*), or * for every Kibana API. Entries are not checked against the known APIs, and one that matches no API has no effect.

        Not more than 100 elements. Maximum length of each is 256.

  • connector_ids array[string]

    Array of connector IDs to associate with the agent.

    Not more than 100 elements.

  • enable_elastic_capabilities boolean

    When true, enables built-in Elastic capabilities for the agent.

  • instructions string

    Optional system instructions that define the agent behavior.

  • plugin_ids array[string]

    Array of plugin IDs to assign to the agent.

    Not more than 100 elements.

  • post_execution_workflow_ids array[string]

    Optional list of workflow IDs. When set, these workflows run after the agent finishes each round.

    Not more than 100 elements. Maximum length of each is 512.

  • skill_ids array[string]

    Array of skill IDs to be available to the agent.

    Not more than 100 elements.

  • subagent_ids array[string] Generally available

    Allowlist of subagent IDs this agent may spawn. Missing or empty disables the run_subagent tool. Use '_self' to enable self-fork.

    Not more than 50 elements. Maximum length of each is 64.

  • tools array[object] Required

    Tool selection configuration for the agent.

    Hide tools attribute Show tools attribute object

    Tool selection configuration for the agent.

    • tool_ids array[string] Required

      Array of tool IDs that the agent can use.

  • workflow_ids array[string]

    Optional list of workflow IDs. When set, these workflows run before every agent execution, in order.

    Not more than 100 elements.

• description string Required

  Description of what the agent does.

• id string Required

  Unique identifier for the agent.

• labels array[string]

  Optional labels for categorizing and organizing agents.

• name string Required

  Display name for the agent.