Agents streaming events

For the complete documentation index, see llms.txt. Markdown versions of documentation pages are available by appending .md to the page URL.

agent.session.environment.ready

Emitted when a hosted session environment is ready to connect.

Schema

Schema name: SessionEventAgentSessionEnvironmentReady

• environment: AgentSessionEnvironmentState

  The current environment state.

  • id: string

    The public ID of the environment.

  • error: object { code, message, type } or null

    The error reported while preparing the environment, if any.

    • code: string

      A machine-readable error code.

    • message: string

      A human-readable error message.

    • type: string

      The error type.

  • status: "pending" or "ready" or "connected" or 2 more

    The environment's connection status.

    • "pending"

      The environment is being prepared.

    • "ready"

      The environment is ready to connect.

    • "connected"

      The environment is connected.

    • "disconnected"

      The environment is disconnected.

    • "failed"

      The environment failed to connect.

  • type: string

    The environment type.

• event_id: string

  The unique ID of the event.

• session_id: string

  The ID of the session associated with the event.

• turn_id: string or null

  The ID of the turn associated with the event, when applicable.

• type: "agent.session.environment.ready"

  The type of the object. Always agent.session.environment.ready.

  • "agent.session.environment.ready"

Example

{
  "type": "agent.session.environment.ready",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "environment": {
    "id": "id",
    "type": "type",
    "status": "pending",
    "error": {
      "type": "type",
      "code": "code",
      "message": "message"
    }
  }
}

agent.session.environment.reset

Emitted after a hosted sandbox is replaced. Conversation history survives; changes to the previous sandbox's files and processes do not.

Schema

Schema name: SessionEventAgentSessionEnvironmentReset

• environment_id: string

  The stable environment ID, retained across sandbox replacements.

• event_id: string

  The unique ID of the event.

• reset_count: number

  Monotonically increasing reset number. Repeated notifications share this number.

• session_id: string

  The ID of the session associated with the event.

• turn_id: string or null

  The associated turn, when applicable.

• type: "agent.session.environment.reset"

  The type of the object. Always agent.session.environment.reset.

  • "agent.session.environment.reset"

Example

{
  "type": "agent.session.environment.reset",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "environment_id": "environment_id",
  "reset_count": 0
}

agent.output.command_execution_output.delta

Emitted when command execution produces an output delta.

Schema

Schema name: SessionEventAgentOutputCommandExecutionOutputDelta

• delta: string

  The output text that was appended.

• event_id: string

  The unique ID of the event.

• item_id: string

  The ID of the command execution item.

• output_index: number

  The index of the item in the turn output.

• session_id: string

  The ID of the session associated with the event.

• turn_id: string or null

  The ID of the turn associated with the event, when applicable.

• type: "agent.output.command_execution_output.delta"

  The type of the object. Always agent.output.command_execution_output.delta.

  • "agent.output.command_execution_output.delta"

Example

{
  "type": "agent.output.command_execution_output.delta",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "item_id": "item_id",
  "output_index": 0,
  "delta": "delta"
}

agent.session.created

Emitted when a session is created.

Schema

Schema name: SessionEventAgentSessionCreated

• event_id: string

  The unique ID of the event.

• session: AgentSession

  The session that was created.

  • id: string

    The ID of the session.

  • agent: object { id, instructions, model, 6 more }

    The agent running in the session.

    • id: string

      The ID of the agent.

    • instructions: string or null

      Custom instructions appended to the agent's default base instructions.

    • model: string

      The model used by the agent.

    • multi_agent: MultiAgentConfig

      Configuration for creating and coordinating subagents.

      • enabled: boolean

        Whether subagent tools are enabled. Defaults to false.

      • max_concurrent_subagents: number or null

        Maximum number of subagents that may run concurrently, or null when disabled. Defaults to 6 when enabled.

    • name: string or null

      The reusable agent's name when the session was created, or null if no name was saved. Later changes to the agent's name do not affect this value.

    • reasoning: AgentReasoning

      The agent's reasoning configuration.

      • effort: "none" or "minimal" or "low" or 4 more or null

        The requested reasoning effort, or null when the model selects its own default.

        • "none"

        • "minimal"

        • "low"

        • "medium"

        • "high"

        • "xhigh"

        • "max"

      • summary: "concise" or "detailed" or "auto" or null

        The requested reasoning summary format, or null when summaries are disabled.

        • "concise"

          Returns a concise reasoning summary when supported.

        • "detailed"

          Returns a detailed reasoning summary when supported.

        • "auto"

          Automatically selects the most detailed summary supported by the model.

    • service_tier: "auto" or "default" or "flex" or 3 more

      The effective service-tier policy for model requests. Defaults to auto.

      • "auto"

      • "default"

      • "flex"

      • "priority"

      • "fast"

      • "ultrafast"

    • text: AgentText

      Configuration for text generated by the agent.

      • format: TextFormat

        The effective output format. Defaults to ordinary text.

        • Text object { type }

          Generates ordinary text without a structured-output constraint.

          • type: "text"

            The type of the object. Always text.

            • "text"

        • JSONSchema object { schema, type }

          Constrains generated text to a JSON Schema.

          • schema: map[unknown]

            The JSON Schema that generated text must match.

          • type: "json_schema"

            The type of the object. Always json_schema.

            • "json_schema"

      • verbosity: "low" or "medium" or "high"

        The amount of text produced by the agent. Defaults to medium.

        • "low"

        • "medium"

        • "high"

    • tools: array of AgentTool

      Tools available to the agent.

      • Function object { defer_loading, description, name, 2 more }

        A function defined by the application.

        • defer_loading: boolean

          Whether the function is deferred and discovered through tool search.

        • description: string

          A description of what the function does.

        • name: string

          The name of the function.

        • parameters: map[unknown]

          A JSON Schema object describing the function's arguments.

        • type: "function"

          The type of the object. Always function.

          • "function"

      • ProgrammaticToolCalling object { enabled, type }

        Enables calling tools from model-generated code.

        • enabled: boolean

          Whether tools can be called from model-generated code.

        • type: "programmatic_tool_calling"

          The type of the object. Always programmatic_tool_calling.

          • "programmatic_tool_calling"

      • Mcp object { allowed_tools, connection_origin, credential_id, 5 more }

        Tools provided by a remote MCP server.

        • allowed_tools: array of string or null

          The MCP tools the agent may call.

        • connection_origin: "service" or "environment"

          Where outbound MCP HTTP connections originate.

          • "service"

          • "environment"

        • credential_id: string or null

          The attached vault credential selected for this MCP server, if any. Optional when exactly one attached credential matches the server URL.

        • request_metadata: map[unknown]

          Metadata included with requests to this MCP server.

        • required: boolean

          Whether this MCP server must initialize before the first turn.

        • server_label: string

          A label used to identify the MCP server in tool calls.

        • transport: McpTransport

          The transport used to connect to the MCP server.

          • HTTP object { server_url, type }

            Connects to an MCP server over HTTP.

            • server_url: string

              The URL of the MCP server.

            • type: "http"

              The type of the object. Always http.

              • "http"

          • Stdio object { args, command, cwd, 2 more }

            Starts an MCP server as a local process.

            • args: array of string

              Arguments passed to the MCP server command.

            • command: string

              The command used to start the MCP server.

            • cwd: string

              The working directory used to start the MCP server.

            • env_vars: array of string

              Environment variable names inherited from the execution environment.

            • type: "stdio"

              The type of the object. Always stdio.

              • "stdio"

        • type: "mcp"

          The type of the object. Always mcp.

          • "mcp"

      • WebSearch object { allowed_domains, context_size, location, 2 more }

        Web search.

        • allowed_domains: array of string or null

          Allowed search domains, or null when the search is unrestricted.

        • context_size: "low" or "medium" or "high"

          The amount of search context made available to the model. Defaults to medium.

          • "low"

          • "medium"

          • "high"

        • location: object { city, country, region, timezone } or null

          Approximate location used to localize search results, if provided.

          • city: string or null

            The city name.

          • country: string or null

            The two-letter ISO country code, such as US.

          • region: string or null

            The region or state name.

          • timezone: string or null

            The IANA timezone, such as America/Los_Angeles.

        • mode: "disabled" or "cached" or "live"

          The source used for web search results.

          • "disabled"

          • "cached"

          • "live"

        • type: "web_search"

          The type of the object. Always web_search.

          • "web_search"

      • ComputerUse object { include_screenshots, type }

        Browser use in an OpenAI-hosted session.

        • include_screenshots: boolean

          Whether computer tool outputs include screenshots.

        • type: "computer_use"

          The type of the object. Always computer_use.

          • "computer_use"

  • created_at: number

    The Unix timestamp, in seconds, when the session was created.

  • environment: Environment

    The execution environment for the session.

    • None object { type }

      The session talks to CCA without selecting or provisioning an execution environment.

      • type: "none"

        The type of the object. Always none.

        • "none"

    • OpenAIHosted object { id, capability_directories, desktop, 7 more }

      An environment hosted by OpenAI.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • desktop: object { enabled }

        The effective desktop configuration.

        • enabled: boolean

          Whether the environment provisions a desktop and browser proxy.

      • files: array of HostedEnvironmentFile

        Files available in the environment, excluding their contents.

        • HostedEnvironmentFileID object { id, file_id, path, 2 more }

          A file copied from the OpenAI Files API.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • file_id: string

            The ID of the uploaded file.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "file_id"

            The type of the object. Always file_id.

            • "file_id"

        • Inline object { id, path, size_bytes, type }

          A file supplied inline when the session was created.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • network: object { access, allowed_domains }

        The effective network access policy for the environment.

        • access: "enabled" or "disabled" or "restricted"

          The environment's network access mode.

          • "enabled"

            Allows unrestricted network access.

          • "disabled"

            Disables network access.

          • "restricted"

            Applies the configured domain restrictions.

        • allowed_domains: array of string

          Domains the environment may access when network access is restricted.

      • packages: object { npm, python, system }

        Packages installed in the environment.

        • npm: array of string

          npm packages installed globally in the environment.

        • python: array of string

          Python packages installed in the environment.

        • system: array of string

          System packages installed in the environment.

      • plugins: array of HostedPlugin

        Plugins installed in the environment, excluding their archive contents.

        • description: string

          The installed plugin description.

        • name: string

          The installed plugin name.

        • type: "inline"

          The type of the object. Always inline.

          • "inline"

      • skills: array of HostedSkill

        Skills installed in the environment, excluding their archive contents.

        • HostedSkillReference object { description, name, skill_id, 2 more }

          A skill installed from the Skills API.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • skill_id: string

            The referenced skill ID.

          • type: "skill_reference"

            The type of the object. Always skill_reference.

            • "skill_reference"

          • version: string

            The concrete skill version installed for this session.

        • Inline object { description, name, type }

          A skill installed from an inline ZIP archive.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • type: "openai_hosted"

        The type of the object. Always openai_hosted.

        • "openai_hosted"

      • container_size: optional "small" or "medium" or "large" or null

        The effective CPU and memory tier, or null when unknown or outside the public tiers.

        • "small"

        • "medium"

        • "large"

    • SelfHosted object { id, capability_directories, remote_url, 2 more }

      An environment hosted by the application.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • remote_url: string

        Pass this URL unchanged to codex exec-server --remote when connecting this environment.

      • type: "self_hosted"

        The type of the object. Always self_hosted.

        • "self_hosted"

      • workspace_directory: string

        The absolute project directory inside the environment. Defaults to /workspace.

  • error: string or null

    The error that caused the session to fail, if any.

  • last_active_at: number

    The Unix timestamp, in seconds, when the session was last active.

  • metadata: map[string]

    Custom string key-value pairs attached to the session.

  • object: "agent.session"

    The object type. Always agent.session.

    • "agent.session"

  • required_actions: array of object { request, request_id, turn_id, type } or object { arguments, call_id, name, 2 more } or object { environment_id, type }

    Actions that must be completed before the session can continue.

    • ComputerUseApprovalRequest object { request, request_id, turn_id, type }

      Respond to a computer-use request.

      • request: object { credential_origin, fields, options, 2 more } or object { origin, reason, type }

        The information needed to render the request.

        • BrowserAuthentication object { credential_origin, fields, options, 2 more }

          A registered form awaiting the application's response.

          • credential_origin: string or null

            The registered form or frame origin where values will be entered.

          • fields: array of object { id, label, required, type }

            Controls to render. All submitted values are sensitive.

            • id: string

              The field ID to submit as field_id in a fields entry.

            • label: string

              The label to display beside the control.

            • required: boolean

              Whether this control requires a nonempty value.

            • type: string

              The rendering type, such as email, password, or text.

          • options: array of object { id, field_ids, label }

            Sign-in methods. Empty for a plain form.

            • id: string

              The option ID to submit as selected_option.

            • field_ids: array of string

              IDs from the registered fields that this method accepts.

            • label: string

              The method label to display.

          • reason: string or null

            Why the agent needs the user to sign in.

          • type: "browser_authentication"

            The type of the object. Always browser_authentication.

            • "browser_authentication"

        • BrowserOriginAccess object { origin, reason, type }

          A browser origin awaiting the application's approval decision.

          • origin: string

            The origin the browser needs permission to access.

          • reason: string or null

            The browser's explanation for this request, or null when unavailable.

          • type: "browser_origin_access"

            The type of the object. Always browser_origin_access.

            • "browser_origin_access"

      • request_id: string

        The registered request ID to echo when responding.

      • turn_id: string

        The turn that requested approval.

      • type: "computer_use_approval_request"

        The type of the object. Always computer_use_approval_request.

        • "computer_use_approval_request"

    • FunctionCall object { arguments, call_id, name, 2 more }

      Run a function tool and submit its result.

      • arguments: unknown

        The arguments supplied by the model.

      • call_id: string

        The ID to include when submitting the function result.

      • name: string

        The function name.

      • turn_id: string

        The ID of the turn that requested the function call.

      • type: "function_call"

        The type of the object. Always function_call.

        • "function_call"

    • EnvironmentConnection object { environment_id, type }

      Reconnect a session environment.

      • environment_id: string

        The ID of the environment to reconnect.

      • type: "environment_connection"

        The type of the object. Always environment_connection.

        • "environment_connection"

  • status: "idle" or "in_progress" or "requires_action" or "failed"

    The current status of the session.

    • "idle"

      The session has no turn in progress and is ready for input. A hosted environment may still be provisioning.

    • "in_progress"

      The session is processing a turn.

    • "requires_action"

      The session is waiting for one or more required actions.

    • "failed"

      The session failed.

  • usage: TokenUsage or null

    Best-effort token usage for the session, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

  • vault_ids: array of string

    The IDs of vaults made available to the session.

• type: "agent.session.created"

  The type of the object. Always agent.session.created.

  • "agent.session.created"

Example

{
  "type": "agent.session.created",
  "event_id": "event_id",
  "session": {
    "metadata": {
      "foo": "string"
    },
    "id": "id",
    "object": "agent.session",
    "created_at": 0,
    "last_active_at": 0,
    "status": "idle",
    "required_actions": [
      {
        "type": "computer_use_approval_request",
        "turn_id": "turn_id",
        "request_id": "request_id",
        "request": {
          "type": "browser_authentication",
          "reason": "reason",
          "credential_origin": "credential_origin",
          "fields": [
            {
              "id": "id",
              "label": "label",
              "type": "type",
              "required": true
            }
          ],
          "options": [
            {
              "id": "id",
              "label": "label",
              "field_ids": [
                "string"
              ]
            }
          ]
        }
      }
    ],
    "error": "error",
    "agent": {
      "id": "id",
      "name": "name",
      "model": "model",
      "reasoning": {
        "effort": "none",
        "summary": "concise"
      },
      "text": {
        "format": {
          "type": "text"
        },
        "verbosity": "low"
      },
      "service_tier": "auto",
      "instructions": "instructions",
      "tools": [
        {
          "type": "function",
          "name": "name",
          "description": "description",
          "parameters": {
            "foo": "bar"
          },
          "defer_loading": true
        }
      ],
      "multi_agent": {
        "enabled": true,
        "max_concurrent_subagents": 1
      }
    },
    "environment": {
      "type": "none"
    },
    "vault_ids": [
      "string"
    ],
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  }
}

agent.session.turn.created

Emitted when a turn is created.

Schema

Schema name: SessionEventAgentSessionTurnCreated

• event_id: string

  The unique ID of the event.

• session_id: string

  The ID of the session associated with the event.

• turn: Turn

  The turn at the time it was created.

  • id: string

    The ID of the turn.

  • agent_id: string

    The ID of the agent that ran the turn.

  • completed_at: number or null

    The Unix timestamp, in seconds, when the turn reached a terminal state.

  • created_at: number

    The Unix timestamp, in seconds, used to order the turn by creation time. Subagent turns use their start time, falling back to completion time or the subagent opening time when the preceding timestamps are unavailable.

  • error: SessionTurnError or null

    A customer-safe error. Non-null only for a failed turn.

    • code: "context_length_exceeded" or "session_budget_exceeded" or "usage_limit_exceeded" or 20 more

      A stable, machine-readable failure category.

      • "context_length_exceeded"

        The request exceeds the model's context window.

      • "session_budget_exceeded"

        The session has reached its usage budget.

      • "usage_limit_exceeded"

        The organization has reached a usage, plan, or billing limit.

      • "project_spend_limit_exceeded"

        The project has reached its enforced spend limit.

      • "organization_spend_limit_exceeded"

        The organization has reached its enforced spend limit.

      • "organization_usage_limit_exceeded"

        The organization has reached its OpenAI-assigned usage limit.

      • "billing_not_active"

        Billing is not active for the account.

      • "credit_balance_exhausted"

        The organization has no API credits remaining.

      • "rate_limit_exceeded"

        The request exceeds the available rate limit.

      • "flex_unavailable"

        Flex processing is temporarily unavailable.

      • "server_overloaded"

        The model service is temporarily overloaded.

      • "cyber_policy"

        The request was rejected by a safety policy.

      • "misalignment_policy_violation"

        The request was blocked by the safety systems.

      • "connection_failed"

        The request could not connect to the model service.

      • "server_error"

        The model service encountered an unexpected error.

      • "authentication_error"

        The API credentials are invalid or lack the required access.

      • "invalid_request"

        The request contains invalid input or configuration.

      • "resource_not_found"

        The requested model or resource is unavailable.

      • "sandbox_error"

        The request could not complete in its execution environment.

      • "executor_version_incompatible"

        The executor must be upgraded before it can run this turn.

      • "active_turn_not_steerable"

        The session cannot accept additional input while a request is running.

      • "request_timeout"

        The request timed out before the model service responded.

      • "internal_error"

        An unexpected internal error prevented the session request from completing.

    • message: string

      A customer-safe explanation of the failure.

  • object: "agent.session.turn"

    The object type. Always agent.session.turn.

    • "agent.session.turn"

  • session_id: string

    The ID of the session that owns the turn.

  • started_at: number or null

    The Unix timestamp, in seconds, when the turn started.

  • status: "queued" or "in_progress" or "waiting" or 3 more

    The current status of the turn.

    • "queued"

      The turn is waiting to start.

    • "in_progress"

      The turn is in progress.

    • "waiting"

      The turn is waiting for external input.

    • "completed"

      The turn completed successfully.

    • "failed"

      The turn failed.

    • "cancelled"

      The turn was cancelled.

  • subagent_id: string or null

    The ID of the subagent that ran the turn, if applicable.

  • usage: TokenUsage or null

    Best-effort token usage for the turn, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

• turn_id: string

  The ID of the turn associated with the event.

• type: "agent.session.turn.created"

  The type of the object. Always agent.session.turn.created.

  • "agent.session.turn.created"

Example

{
  "type": "agent.session.turn.created",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "turn": {
    "id": "id",
    "object": "agent.session.turn",
    "session_id": "session_id",
    "agent_id": "agent_id",
    "subagent_id": "subagent_id",
    "status": "queued",
    "created_at": 0,
    "started_at": 0,
    "completed_at": 0,
    "error": {
      "code": "context_length_exceeded",
      "message": "message"
    },
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  }
}

agent.session.turn.in_progress

Emitted when a turn starts running.

Schema

Schema name: SessionEventAgentSessionTurnInProgress

• event_id: string

  The unique ID of the event.

• session_id: string

  The ID of the session associated with the event.

• turn: Turn

  The turn at the time it started running.

  • id: string

    The ID of the turn.

  • agent_id: string

    The ID of the agent that ran the turn.

  • completed_at: number or null

    The Unix timestamp, in seconds, when the turn reached a terminal state.

  • created_at: number

    The Unix timestamp, in seconds, used to order the turn by creation time. Subagent turns use their start time, falling back to completion time or the subagent opening time when the preceding timestamps are unavailable.

  • error: SessionTurnError or null

    A customer-safe error. Non-null only for a failed turn.

    • code: "context_length_exceeded" or "session_budget_exceeded" or "usage_limit_exceeded" or 20 more

      A stable, machine-readable failure category.

      • "context_length_exceeded"

        The request exceeds the model's context window.

      • "session_budget_exceeded"

        The session has reached its usage budget.

      • "usage_limit_exceeded"

        The organization has reached a usage, plan, or billing limit.

      • "project_spend_limit_exceeded"

        The project has reached its enforced spend limit.

      • "organization_spend_limit_exceeded"

        The organization has reached its enforced spend limit.

      • "organization_usage_limit_exceeded"

        The organization has reached its OpenAI-assigned usage limit.

      • "billing_not_active"

        Billing is not active for the account.

      • "credit_balance_exhausted"

        The organization has no API credits remaining.

      • "rate_limit_exceeded"

        The request exceeds the available rate limit.

      • "flex_unavailable"

        Flex processing is temporarily unavailable.

      • "server_overloaded"

        The model service is temporarily overloaded.

      • "cyber_policy"

        The request was rejected by a safety policy.

      • "misalignment_policy_violation"

        The request was blocked by the safety systems.

      • "connection_failed"

        The request could not connect to the model service.

      • "server_error"

        The model service encountered an unexpected error.

      • "authentication_error"

        The API credentials are invalid or lack the required access.

      • "invalid_request"

        The request contains invalid input or configuration.

      • "resource_not_found"

        The requested model or resource is unavailable.

      • "sandbox_error"

        The request could not complete in its execution environment.

      • "executor_version_incompatible"

        The executor must be upgraded before it can run this turn.

      • "active_turn_not_steerable"

        The session cannot accept additional input while a request is running.

      • "request_timeout"

        The request timed out before the model service responded.

      • "internal_error"

        An unexpected internal error prevented the session request from completing.

    • message: string

      A customer-safe explanation of the failure.

  • object: "agent.session.turn"

    The object type. Always agent.session.turn.

    • "agent.session.turn"

  • session_id: string

    The ID of the session that owns the turn.

  • started_at: number or null

    The Unix timestamp, in seconds, when the turn started.

  • status: "queued" or "in_progress" or "waiting" or 3 more

    The current status of the turn.

    • "queued"

      The turn is waiting to start.

    • "in_progress"

      The turn is in progress.

    • "waiting"

      The turn is waiting for external input.

    • "completed"

      The turn completed successfully.

    • "failed"

      The turn failed.

    • "cancelled"

      The turn was cancelled.

  • subagent_id: string or null

    The ID of the subagent that ran the turn, if applicable.

  • usage: TokenUsage or null

    Best-effort token usage for the turn, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

• turn_id: string

  The ID of the turn associated with the event.

• type: "agent.session.turn.in_progress"

  The type of the object. Always agent.session.turn.in_progress.

  • "agent.session.turn.in_progress"

Example

{
  "type": "agent.session.turn.in_progress",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "turn": {
    "id": "id",
    "object": "agent.session.turn",
    "session_id": "session_id",
    "agent_id": "agent_id",
    "subagent_id": "subagent_id",
    "status": "queued",
    "created_at": 0,
    "started_at": 0,
    "completed_at": 0,
    "error": {
      "code": "context_length_exceeded",
      "message": "message"
    },
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  }
}

agent.session.turn.completed

Emitted when a turn completes.

Schema

Schema name: SessionEventAgentSessionTurnCompleted

• event_id: string

  The unique ID of the event.

• session_id: string

  The ID of the session associated with the event.

• turn: Turn

  The completed turn.

  • id: string

    The ID of the turn.

  • agent_id: string

    The ID of the agent that ran the turn.

  • completed_at: number or null

    The Unix timestamp, in seconds, when the turn reached a terminal state.

  • created_at: number

    The Unix timestamp, in seconds, used to order the turn by creation time. Subagent turns use their start time, falling back to completion time or the subagent opening time when the preceding timestamps are unavailable.

  • error: SessionTurnError or null

    A customer-safe error. Non-null only for a failed turn.

    • code: "context_length_exceeded" or "session_budget_exceeded" or "usage_limit_exceeded" or 20 more

      A stable, machine-readable failure category.

      • "context_length_exceeded"

        The request exceeds the model's context window.

      • "session_budget_exceeded"

        The session has reached its usage budget.

      • "usage_limit_exceeded"

        The organization has reached a usage, plan, or billing limit.

      • "project_spend_limit_exceeded"

        The project has reached its enforced spend limit.

      • "organization_spend_limit_exceeded"

        The organization has reached its enforced spend limit.

      • "organization_usage_limit_exceeded"

        The organization has reached its OpenAI-assigned usage limit.

      • "billing_not_active"

        Billing is not active for the account.

      • "credit_balance_exhausted"

        The organization has no API credits remaining.

      • "rate_limit_exceeded"

        The request exceeds the available rate limit.

      • "flex_unavailable"

        Flex processing is temporarily unavailable.

      • "server_overloaded"

        The model service is temporarily overloaded.

      • "cyber_policy"

        The request was rejected by a safety policy.

      • "misalignment_policy_violation"

        The request was blocked by the safety systems.

      • "connection_failed"

        The request could not connect to the model service.

      • "server_error"

        The model service encountered an unexpected error.

      • "authentication_error"

        The API credentials are invalid or lack the required access.

      • "invalid_request"

        The request contains invalid input or configuration.

      • "resource_not_found"

        The requested model or resource is unavailable.

      • "sandbox_error"

        The request could not complete in its execution environment.

      • "executor_version_incompatible"

        The executor must be upgraded before it can run this turn.

      • "active_turn_not_steerable"

        The session cannot accept additional input while a request is running.

      • "request_timeout"

        The request timed out before the model service responded.

      • "internal_error"

        An unexpected internal error prevented the session request from completing.

    • message: string

      A customer-safe explanation of the failure.

  • object: "agent.session.turn"

    The object type. Always agent.session.turn.

    • "agent.session.turn"

  • session_id: string

    The ID of the session that owns the turn.

  • started_at: number or null

    The Unix timestamp, in seconds, when the turn started.

  • status: "queued" or "in_progress" or "waiting" or 3 more

    The current status of the turn.

    • "queued"

      The turn is waiting to start.

    • "in_progress"

      The turn is in progress.

    • "waiting"

      The turn is waiting for external input.

    • "completed"

      The turn completed successfully.

    • "failed"

      The turn failed.

    • "cancelled"

      The turn was cancelled.

  • subagent_id: string or null

    The ID of the subagent that ran the turn, if applicable.

  • usage: TokenUsage or null

    Best-effort token usage for the turn, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

• turn_id: string

  The ID of the turn associated with the event.

• type: "agent.session.turn.completed"

  The type of the object. Always agent.session.turn.completed.

  • "agent.session.turn.completed"

• usage: TokenUsage or null

  Token usage by the root agent during the turn, when available.

Example

{
  "type": "agent.session.turn.completed",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "turn": {
    "id": "id",
    "object": "agent.session.turn",
    "session_id": "session_id",
    "agent_id": "agent_id",
    "subagent_id": "subagent_id",
    "status": "queued",
    "created_at": 0,
    "started_at": 0,
    "completed_at": 0,
    "error": {
      "code": "context_length_exceeded",
      "message": "message"
    },
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  },
  "usage": {
    "input_tokens": 0,
    "input_tokens_details": {
      "cached_tokens": 0
    },
    "output_tokens": 0,
    "output_tokens_details": {
      "reasoning_tokens": 0
    },
    "total_tokens": 0
  }
}

agent.session.turn.failed

Emitted when a turn fails.

Schema

Schema name: SessionEventAgentSessionTurnFailed

• event_id: string

  The unique ID of the event.

• session_id: string

  The ID of the session associated with the event.

• turn: Turn

  The failed turn.

  • id: string

    The ID of the turn.

  • agent_id: string

    The ID of the agent that ran the turn.

  • completed_at: number or null

    The Unix timestamp, in seconds, when the turn reached a terminal state.

  • created_at: number

    The Unix timestamp, in seconds, used to order the turn by creation time. Subagent turns use their start time, falling back to completion time or the subagent opening time when the preceding timestamps are unavailable.

  • error: SessionTurnError or null

    A customer-safe error. Non-null only for a failed turn.

    • code: "context_length_exceeded" or "session_budget_exceeded" or "usage_limit_exceeded" or 20 more

      A stable, machine-readable failure category.

      • "context_length_exceeded"

        The request exceeds the model's context window.

      • "session_budget_exceeded"

        The session has reached its usage budget.

      • "usage_limit_exceeded"

        The organization has reached a usage, plan, or billing limit.

      • "project_spend_limit_exceeded"

        The project has reached its enforced spend limit.

      • "organization_spend_limit_exceeded"

        The organization has reached its enforced spend limit.

      • "organization_usage_limit_exceeded"

        The organization has reached its OpenAI-assigned usage limit.

      • "billing_not_active"

        Billing is not active for the account.

      • "credit_balance_exhausted"

        The organization has no API credits remaining.

      • "rate_limit_exceeded"

        The request exceeds the available rate limit.

      • "flex_unavailable"

        Flex processing is temporarily unavailable.

      • "server_overloaded"

        The model service is temporarily overloaded.

      • "cyber_policy"

        The request was rejected by a safety policy.

      • "misalignment_policy_violation"

        The request was blocked by the safety systems.

      • "connection_failed"

        The request could not connect to the model service.

      • "server_error"

        The model service encountered an unexpected error.

      • "authentication_error"

        The API credentials are invalid or lack the required access.

      • "invalid_request"

        The request contains invalid input or configuration.

      • "resource_not_found"

        The requested model or resource is unavailable.

      • "sandbox_error"

        The request could not complete in its execution environment.

      • "executor_version_incompatible"

        The executor must be upgraded before it can run this turn.

      • "active_turn_not_steerable"

        The session cannot accept additional input while a request is running.

      • "request_timeout"

        The request timed out before the model service responded.

      • "internal_error"

        An unexpected internal error prevented the session request from completing.

    • message: string

      A customer-safe explanation of the failure.

  • object: "agent.session.turn"

    The object type. Always agent.session.turn.

    • "agent.session.turn"

  • session_id: string

    The ID of the session that owns the turn.

  • started_at: number or null

    The Unix timestamp, in seconds, when the turn started.

  • status: "queued" or "in_progress" or "waiting" or 3 more

    The current status of the turn.

    • "queued"

      The turn is waiting to start.

    • "in_progress"

      The turn is in progress.

    • "waiting"

      The turn is waiting for external input.

    • "completed"

      The turn completed successfully.

    • "failed"

      The turn failed.

    • "cancelled"

      The turn was cancelled.

  • subagent_id: string or null

    The ID of the subagent that ran the turn, if applicable.

  • usage: TokenUsage or null

    Best-effort token usage for the turn, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

• turn_id: string

  The ID of the turn associated with the event.

• type: "agent.session.turn.failed"

  The type of the object. Always agent.session.turn.failed.

  • "agent.session.turn.failed"

• usage: TokenUsage or null

  Token usage by the root agent during the turn, when available.

Example

{
  "type": "agent.session.turn.failed",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "turn": {
    "id": "id",
    "object": "agent.session.turn",
    "session_id": "session_id",
    "agent_id": "agent_id",
    "subagent_id": "subagent_id",
    "status": "queued",
    "created_at": 0,
    "started_at": 0,
    "completed_at": 0,
    "error": {
      "code": "context_length_exceeded",
      "message": "message"
    },
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  },
  "usage": {
    "input_tokens": 0,
    "input_tokens_details": {
      "cached_tokens": 0
    },
    "output_tokens": 0,
    "output_tokens_details": {
      "reasoning_tokens": 0
    },
    "total_tokens": 0
  }
}

agent.session.turn.cancelled

Emitted when a turn is cancelled.

Schema

Schema name: SessionEventAgentSessionTurnCancelled

• event_id: string

  The unique ID of the event.

• session_id: string

  The ID of the session associated with the event.

• turn: Turn

  The cancelled turn.

  • id: string

    The ID of the turn.

  • agent_id: string

    The ID of the agent that ran the turn.

  • completed_at: number or null

    The Unix timestamp, in seconds, when the turn reached a terminal state.

  • created_at: number

    The Unix timestamp, in seconds, used to order the turn by creation time. Subagent turns use their start time, falling back to completion time or the subagent opening time when the preceding timestamps are unavailable.

  • error: SessionTurnError or null

    A customer-safe error. Non-null only for a failed turn.

    • code: "context_length_exceeded" or "session_budget_exceeded" or "usage_limit_exceeded" or 20 more

      A stable, machine-readable failure category.

      • "context_length_exceeded"

        The request exceeds the model's context window.

      • "session_budget_exceeded"

        The session has reached its usage budget.

      • "usage_limit_exceeded"

        The organization has reached a usage, plan, or billing limit.

      • "project_spend_limit_exceeded"

        The project has reached its enforced spend limit.

      • "organization_spend_limit_exceeded"

        The organization has reached its enforced spend limit.

      • "organization_usage_limit_exceeded"

        The organization has reached its OpenAI-assigned usage limit.

      • "billing_not_active"

        Billing is not active for the account.

      • "credit_balance_exhausted"

        The organization has no API credits remaining.

      • "rate_limit_exceeded"

        The request exceeds the available rate limit.

      • "flex_unavailable"

        Flex processing is temporarily unavailable.

      • "server_overloaded"

        The model service is temporarily overloaded.

      • "cyber_policy"

        The request was rejected by a safety policy.

      • "misalignment_policy_violation"

        The request was blocked by the safety systems.

      • "connection_failed"

        The request could not connect to the model service.

      • "server_error"

        The model service encountered an unexpected error.

      • "authentication_error"

        The API credentials are invalid or lack the required access.

      • "invalid_request"

        The request contains invalid input or configuration.

      • "resource_not_found"

        The requested model or resource is unavailable.

      • "sandbox_error"

        The request could not complete in its execution environment.

      • "executor_version_incompatible"

        The executor must be upgraded before it can run this turn.

      • "active_turn_not_steerable"

        The session cannot accept additional input while a request is running.

      • "request_timeout"

        The request timed out before the model service responded.

      • "internal_error"

        An unexpected internal error prevented the session request from completing.

    • message: string

      A customer-safe explanation of the failure.

  • object: "agent.session.turn"

    The object type. Always agent.session.turn.

    • "agent.session.turn"

  • session_id: string

    The ID of the session that owns the turn.

  • started_at: number or null

    The Unix timestamp, in seconds, when the turn started.

  • status: "queued" or "in_progress" or "waiting" or 3 more

    The current status of the turn.

    • "queued"

      The turn is waiting to start.

    • "in_progress"

      The turn is in progress.

    • "waiting"

      The turn is waiting for external input.

    • "completed"

      The turn completed successfully.

    • "failed"

      The turn failed.

    • "cancelled"

      The turn was cancelled.

  • subagent_id: string or null

    The ID of the subagent that ran the turn, if applicable.

  • usage: TokenUsage or null

    Best-effort token usage for the turn, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

• turn_id: string

  The ID of the turn associated with the event.

• type: "agent.session.turn.cancelled"

  The type of the object. Always agent.session.turn.cancelled.

  • "agent.session.turn.cancelled"

• usage: TokenUsage or null

  Token usage by the root agent during the turn, when available.

Example

{
  "type": "agent.session.turn.cancelled",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "turn": {
    "id": "id",
    "object": "agent.session.turn",
    "session_id": "session_id",
    "agent_id": "agent_id",
    "subagent_id": "subagent_id",
    "status": "queued",
    "created_at": 0,
    "started_at": 0,
    "completed_at": 0,
    "error": {
      "code": "context_length_exceeded",
      "message": "message"
    },
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  },
  "usage": {
    "input_tokens": 0,
    "input_tokens_details": {
      "cached_tokens": 0
    },
    "output_tokens": 0,
    "output_tokens_details": {
      "reasoning_tokens": 0
    },
    "total_tokens": 0
  }
}

agent.session.turn.item.added

Emitted when an item is added to a turn.

Schema

Schema name: SessionEventAgentSessionTurnItemAdded

• event_id: string

  The unique ID of the event.

• item: AgentSessionItem

  The item that was added.

  • AgentSessionMessage object { id, content, phase, 4 more }

    A user or assistant message recorded in a session.

    • id: string or null

      The ID of this item, or null for legacy user messages whose ID was not recorded.

    • content: array of AgentSessionMessageContent

      The content of the message. User messages contain input text or images; assistant messages contain output text.

      • InputText object { text, type }

        Text supplied by the user.

        • text: string

          The text supplied by the user.

        • type: "input_text"

          The type of the object. Always input_text.

          • "input_text"

      • InputImage object { image_url, type }

        An image supplied by the user.

        • image_url: string

          The URL of the image supplied by the user, which may be a base64-encoded data URL.

        • type: "input_image"

          The type of the object. Always input_image.

          • "input_image"

      • OutputText object { text, type }

        Text produced by the assistant.

        • text: string

          The text produced by the assistant.

        • type: "output_text"

          The type of the object. Always output_text.

          • "output_text"

    • phase: "commentary" or "final_answer" or null

      The phase of an assistant message. Null for user messages.

      • "commentary"

        Commentary produced while the agent works.

      • "final_answer"

        The agent's final answer.

    • role: "user" or "assistant"

      The role of the message author.

      • "user"

      • "assistant"

    • status: AgentOutputItemStatus

      The status of the message. User messages are always completed.

      • "in_progress"

        The item is in progress.

      • "completed"

        The item is complete.

      • "incomplete"

        The item stopped before completing.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "message"

      The item type. Always message.

      • "message"

  • AgentReasoningItem object { id, status, summary, 2 more }

    A reasoning item produced by the agent.

    • id: string

      The ID of the reasoning item.

    • status: AgentOutputItemStatus or null

      The status of the reasoning item.

    • summary: array of SummaryText

      The reasoning summaries produced by the agent.

      • text: string

        The reasoning summary text.

      • type: "summary_text"

        The content type. Always summary_text.

        • "summary_text"

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "reasoning"

      The item type. Always reasoning.

      • "reasoning"

  • AgentFunctionCallItem object { id, arguments, call_id, 4 more }

    A function call produced by the agent.

    • id: string

      The ID of the function call item.

    • arguments: unknown

      The arguments to pass to the function.

    • call_id: string

      The ID used to submit the function result.

    • name: string

      The name of the function to call.

    • status: AgentFunctionCallStatus

      The status of the function call.

      • "in_progress"

        The call is in progress.

      • "completed"

        The call completed successfully.

      • "failed"

        The call failed.

      • "incomplete"

        The call stopped before completing.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "function_call"

      The item type. Always function_call.

      • "function_call"

  • FunctionCallOutput object { id, call_id, error, 4 more }

    The result supplied for a function call.

    • id: string

      The ID of the function call output item.

    • call_id: string

      The ID of the function call that produced this output.

    • error: string or null

      The error message, if the call failed.

    • output: AgentFunctionCallOutput or null

      The function result, if the call succeeded.

      • string

      • array of InputContent

        • InputText object { text, type }

          Text input recorded in a session item.

          • text: string

            The text supplied to the agent.

          • type: "input_text"

            The type of the object. Always input_text.

            • "input_text"

        • InputImage object { image_url, type }

          Image input recorded in a session item.

          • image_url: string

            The URL of the image supplied to the agent, which may be a base64-encoded data URL.

          • type: "input_image"

            The type of the object. Always input_image.

            • "input_image"

    • status: AgentFunctionCallStatus

      The status of the function call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "function_call_output"

      The item type. Always function_call_output.

      • "function_call_output"

  • AgentMessage object { id, content, recipient_agent_id, 3 more }

    A message exchanged between agent threads.

    • id: string

      The ID of the message.

    • content: array of AgentContent

      The content exchanged between the agents.

      • OutputText object { text, type }

        A text content part produced by the agent.

        • text: string

          The text produced by the agent.

        • type: "output_text"

          The content type. Always output_text.

          • "output_text"

      • EncryptedContent object { encrypted_content, type }

        Encrypted content exchanged between agents.

        • encrypted_content: string

          The encrypted content payload.

        • type: "encrypted_content"

          The content type. Always encrypted_content.

          • "encrypted_content"

    • recipient_agent_id: string

      The ID or name of the receiving agent.

    • sender_agent_id: string

      The ID or name of the sending agent.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "agent_message"

      The item type. Always agent_message.

      • "agent_message"

  • AgentMcpCallItem object { id, arguments, error, 6 more }

    A call to a tool on an MCP server.

    • id: string

      The ID of the MCP call item.

    • arguments: unknown

      The arguments passed to the MCP tool.

    • error: unknown

      The error returned by the MCP tool, if any.

    • name: string

      The name of the MCP tool.

    • output: unknown

      The output returned by the MCP tool, if any.

    • server_label: string

      The label of the MCP server.

    • status: AgentFunctionCallStatus

      The status of the MCP tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "mcp_call"

      The item type. Always mcp_call.

      • "mcp_call"

  • ComputerUseCall object { id, output, status, 3 more }

    One execution of the platform-provided computer-use capability.

    • id: string

      The ID of the activity item.

    • output: object { image_url, type } or null

      The last screenshot emitted by the model. Null when screenshot inclusion is disabled or the call emitted no screenshot.

      • image_url: string

        The complete JPEG image as a base64 data URL.

      • type: "computer_screenshot"

        The content type. Always computer_screenshot.

        • "computer_screenshot"

    • status: AgentFunctionCallStatus

      The execution status of the activity.

    • title: string or null

      A model-generated description of the activity, when available.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "computer_use_call"

      The item type. Always computer_use_call.

      • "computer_use_call"

  • ComputerUseApprovalRequest object { id, request, request_id, 2 more }

    A credential-free history record of the emitted login request.

    • id: string

      The stable history item ID.

    • request: object { credential_origin, fields, options, 2 more }

      A registered form awaiting the application's response.

      • credential_origin: string or null

        The registered form or frame origin where values will be entered.

      • fields: array of object { id, label, required, type }

        Controls to render. All submitted values are sensitive.

        • id: string

          The field ID to submit as field_id in a fields entry.

        • label: string

          The label to display beside the control.

        • required: boolean

          Whether this control requires a nonempty value.

        • type: string

          The rendering type, such as email, password, or text.

      • options: array of object { id, field_ids, label }

        Sign-in methods. Empty for a plain form.

        • id: string

          The option ID to submit as selected_option.

        • field_ids: array of string

          IDs from the registered fields that this method accepts.

        • label: string

          The method label to display.

      • reason: string or null

        Why the agent needs the user to sign in.

      • type: "browser_authentication"

        The type of the object. Always browser_authentication.

        • "browser_authentication"

    • request_id: string

    • turn_id: string

    • type: "computer_use_approval_request"

      The item type. Always computer_use_approval_request.

      • "computer_use_approval_request"

  • ComputerUseApprovalRequestResult object { id, request_id, response, 2 more }

    A credential-free record of an admitted response, not proof of completion.

    • id: string

      The stable history item ID.

    • request_id: string

      The registered request answered by this item.

    • response: object { action, selected_option, type } or object { action, type }

      The admitted response, without submitted credential values.

      • Submit object { action, selected_option, type }

        • action: "submit"

          • "submit"

        • selected_option: string or null

          The chosen sign-in method, or null when no options were offered.

        • type: "browser_authentication"

          • "browser_authentication"

      • Cancel object { action, type }

        • action: "cancel"

          • "cancel"

        • type: "browser_authentication"

          • "browser_authentication"

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "computer_use_approval_request_result"

      • "computer_use_approval_request_result"

  • AgentWebSearchCallItem object { id, action, status, 2 more }

    A web search call produced by the agent.

    • id: string

      The ID of the web search call.

    • action: WebSearchAction or null

      The action performed by the web search tool.

      • Search object { queries, query, type }

        A search query or group of search queries.

        • queries: array of string or null

          The search queries, when multiple queries were used.

        • query: string or null

          The search query, when a single query was used.

        • type: "search"

          The type of the object. Always search.

          • "search"

      • OpenPage object { type, url }

        Opens a web page.

        • type: "open_page"

          The type of the object. Always open_page.

          • "open_page"

        • url: string or null

          The URL of the page that was opened.

      • FindInPage object { pattern, type, url }

        Finds text within a web page.

        • pattern: string or null

          The text pattern that was searched for.

        • type: "find_in_page"

          The type of the object. Always find_in_page.

          • "find_in_page"

        • url: string or null

          The URL of the page that was searched.

      • Other object { type }

        Another web search action.

        • type: "other"

          The type of the object. Always other.

          • "other"

    • status: AgentOutputItemStatus

      The status of the web search call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "web_search_call"

      The item type. Always web_search_call.

      • "web_search_call"

  • AgentCommandExecutionItem object { id, command, cwd, 6 more }

    A command execution produced by the agent.

    • id: string

      The ID of the command execution item.

    • command: string

      The command that was executed.

    • cwd: string or null

      The working directory used to execute the command.

    • duration_ms: number or null

      The command duration in milliseconds.

    • exit_code: number or null

      The process exit code, if the command completed.

    • output: string or null

      The command output, if available.

    • status: AgentFunctionCallStatus

      The status of the command execution.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "command_execution"

      The item type. Always command_execution.

      • "command_execution"

  • AgentCreateSubagentCallItem object { id, agent_id, content, 5 more }

    A request to spawn a subagent.

    • id: string

      The ID of the tool call item.

    • agent_id: string

      The ID of the agent that requested the subagent.

    • content: array of AgentContent

      The task given to the spawned agent.

      • OutputText object { text, type }

        A text content part produced by the agent.

      • EncryptedContent object { encrypted_content, type }

        Encrypted content exchanged between agents.

    • model: string or null

      The model requested for the spawned agent.

    • reasoning_effort: string or null

      The reasoning effort requested for the spawned agent.

    • status: AgentFunctionCallStatus

      The status of the tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "create_subagent_call"

      The item type. Always create_subagent_call.

      • "create_subagent_call"

        The current public item type.

  • AgentSendSubagentInputCallItem object { id, content, recipient_agent_id, 4 more }

    A request to send input to another agent.

    • id: string

      The ID of the tool call item.

    • content: array of AgentContent

      The input sent to the receiving agent.

      • OutputText object { text, type }

        A text content part produced by the agent.

      • EncryptedContent object { encrypted_content, type }

        Encrypted content exchanged between agents.

    • recipient_agent_id: string

      The ID of the agent receiving the input.

    • sender_agent_id: string

      The ID of the agent sending the input.

    • status: AgentFunctionCallStatus

      The status of the tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "send_subagent_input_call"

      The item type. Always send_subagent_input_call.

      • "send_subagent_input_call"

        The current public item type.

  • AgentResumeSubagentCallItem object { id, recipient_agent_id, sender_agent_id, 3 more }

    A request to resume a subagent.

    • id: string

      The ID of the tool call item.

    • recipient_agent_id: string

      The ID of the agent to resume.

    • sender_agent_id: string

      The ID of the agent requesting the resume.

    • status: AgentFunctionCallStatus

      The status of the tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "resume_subagent_call"

      The item type. Always resume_subagent_call.

      • "resume_subagent_call"

        The current public item type.

  • AgentWaitForSubagentsCallItem object { id, recipient_agent_ids, sender_agent_id, 3 more }

    A request to wait for one or more subagents.

    • id: string

      The ID of the tool call item.

    • recipient_agent_ids: array of string

      The IDs of the agents to wait for.

    • sender_agent_id: string

      The ID of the agent waiting for results.

    • status: AgentFunctionCallStatus

      The status of the tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "wait_for_subagents_call"

      The item type. Always wait_for_subagents_call.

      • "wait_for_subagents_call"

        The current public item type.

  • AgentInterruptSubagentCallItem object { id, recipient_agent_id, sender_agent_id, 3 more }

    A request to interrupt a subagent's current turn. The subagent remains available.

    • id: string

      The ID of the tool call item.

    • recipient_agent_id: string

      The ID of the agent to interrupt.

    • sender_agent_id: string

      The ID of the agent requesting the interrupt.

    • status: AgentFunctionCallStatus

      The status of the tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "interrupt_subagent_call"

      The item type. Always interrupt_subagent_call.

      • "interrupt_subagent_call"

        The current public item type.

  • AgentCloseSubagentCallItem object { id, recipient_agent_id, sender_agent_id, 3 more }

    A request to close a subagent.

    • id: string

      The ID of the tool call item.

    • recipient_agent_id: string

      The ID of the agent to close.

    • sender_agent_id: string

      The ID of the agent requesting the close.

    • status: AgentFunctionCallStatus

      The status of the tool call.

    • turn_id: string

      The ID of the turn that contains this item.

    • type: "close_subagent_call"

      The item type. Always close_subagent_call.

      • "close_subagent_call"

        The current public item type.

• output_index: number or null

  The index of the item in the turn output, when the item is agent output.

• session_id: string

  The ID of the session associated with the event.

• turn_id: string or null

  The ID of the turn associated with the event, when applicable.

• type: "agent.session.turn.item.added"

  The type of the object. Always agent.session.turn.item.added.

  • "agent.session.turn.item.added"

Example

{
  "type": "agent.session.turn.item.added",
  "event_id": "event_id",
  "session_id": "session_id",
  "turn_id": "turn_id",
  "output_index": 0,
  "item": {
    "type": "message",
    "id": "id",
    "turn_id": "turn_id",
    "role": "user",
    "content": [
      {
        "type": "input_text",
        "text": "text"
      }
    ],
    "status": "in_progress",
    "phase": "commentary"
  }
}

agent.session.idle

Emitted when a session becomes idle.

Schema

Schema name: SessionEventAgentSessionIdle

• event_id: string

  The unique ID of the event.

• session: AgentSession

  The session that became idle.

  • id: string

    The ID of the session.

  • agent: object { id, instructions, model, 6 more }

    The agent running in the session.

    • id: string

      The ID of the agent.

    • instructions: string or null

      Custom instructions appended to the agent's default base instructions.

    • model: string

      The model used by the agent.

    • multi_agent: MultiAgentConfig

      Configuration for creating and coordinating subagents.

      • enabled: boolean

        Whether subagent tools are enabled. Defaults to false.

      • max_concurrent_subagents: number or null

        Maximum number of subagents that may run concurrently, or null when disabled. Defaults to 6 when enabled.

    • name: string or null

      The reusable agent's name when the session was created, or null if no name was saved. Later changes to the agent's name do not affect this value.

    • reasoning: AgentReasoning

      The agent's reasoning configuration.

      • effort: "none" or "minimal" or "low" or 4 more or null

        The requested reasoning effort, or null when the model selects its own default.

        • "none"

        • "minimal"

        • "low"

        • "medium"

        • "high"

        • "xhigh"

        • "max"

      • summary: "concise" or "detailed" or "auto" or null

        The requested reasoning summary format, or null when summaries are disabled.

        • "concise"

          Returns a concise reasoning summary when supported.

        • "detailed"

          Returns a detailed reasoning summary when supported.

        • "auto"

          Automatically selects the most detailed summary supported by the model.

    • service_tier: "auto" or "default" or "flex" or 3 more

      The effective service-tier policy for model requests. Defaults to auto.

      • "auto"

      • "default"

      • "flex"

      • "priority"

      • "fast"

      • "ultrafast"

    • text: AgentText

      Configuration for text generated by the agent.

      • format: TextFormat

        The effective output format. Defaults to ordinary text.

        • Text object { type }

          Generates ordinary text without a structured-output constraint.

          • type: "text"

            The type of the object. Always text.

            • "text"

        • JSONSchema object { schema, type }

          Constrains generated text to a JSON Schema.

          • schema: map[unknown]

            The JSON Schema that generated text must match.

          • type: "json_schema"

            The type of the object. Always json_schema.

            • "json_schema"

      • verbosity: "low" or "medium" or "high"

        The amount of text produced by the agent. Defaults to medium.

        • "low"

        • "medium"

        • "high"

    • tools: array of AgentTool

      Tools available to the agent.

      • Function object { defer_loading, description, name, 2 more }

        A function defined by the application.

        • defer_loading: boolean

          Whether the function is deferred and discovered through tool search.

        • description: string

          A description of what the function does.

        • name: string

          The name of the function.

        • parameters: map[unknown]

          A JSON Schema object describing the function's arguments.

        • type: "function"

          The type of the object. Always function.

          • "function"

      • ProgrammaticToolCalling object { enabled, type }

        Enables calling tools from model-generated code.

        • enabled: boolean

          Whether tools can be called from model-generated code.

        • type: "programmatic_tool_calling"

          The type of the object. Always programmatic_tool_calling.

          • "programmatic_tool_calling"

      • Mcp object { allowed_tools, connection_origin, credential_id, 5 more }

        Tools provided by a remote MCP server.

        • allowed_tools: array of string or null

          The MCP tools the agent may call.

        • connection_origin: "service" or "environment"

          Where outbound MCP HTTP connections originate.

          • "service"

          • "environment"

        • credential_id: string or null

          The attached vault credential selected for this MCP server, if any. Optional when exactly one attached credential matches the server URL.

        • request_metadata: map[unknown]

          Metadata included with requests to this MCP server.

        • required: boolean

          Whether this MCP server must initialize before the first turn.

        • server_label: string

          A label used to identify the MCP server in tool calls.

        • transport: McpTransport

          The transport used to connect to the MCP server.

          • HTTP object { server_url, type }

            Connects to an MCP server over HTTP.

            • server_url: string

              The URL of the MCP server.

            • type: "http"

              The type of the object. Always http.

              • "http"

          • Stdio object { args, command, cwd, 2 more }

            Starts an MCP server as a local process.

            • args: array of string

              Arguments passed to the MCP server command.

            • command: string

              The command used to start the MCP server.

            • cwd: string

              The working directory used to start the MCP server.

            • env_vars: array of string

              Environment variable names inherited from the execution environment.

            • type: "stdio"

              The type of the object. Always stdio.

              • "stdio"

        • type: "mcp"

          The type of the object. Always mcp.

          • "mcp"

      • WebSearch object { allowed_domains, context_size, location, 2 more }

        Web search.

        • allowed_domains: array of string or null

          Allowed search domains, or null when the search is unrestricted.

        • context_size: "low" or "medium" or "high"

          The amount of search context made available to the model. Defaults to medium.

          • "low"

          • "medium"

          • "high"

        • location: object { city, country, region, timezone } or null

          Approximate location used to localize search results, if provided.

          • city: string or null

            The city name.

          • country: string or null

            The two-letter ISO country code, such as US.

          • region: string or null

            The region or state name.

          • timezone: string or null

            The IANA timezone, such as America/Los_Angeles.

        • mode: "disabled" or "cached" or "live"

          The source used for web search results.

          • "disabled"

          • "cached"

          • "live"

        • type: "web_search"

          The type of the object. Always web_search.

          • "web_search"

      • ComputerUse object { include_screenshots, type }

        Browser use in an OpenAI-hosted session.

        • include_screenshots: boolean

          Whether computer tool outputs include screenshots.

        • type: "computer_use"

          The type of the object. Always computer_use.

          • "computer_use"

  • created_at: number

    The Unix timestamp, in seconds, when the session was created.

  • environment: Environment

    The execution environment for the session.

    • None object { type }

      The session talks to CCA without selecting or provisioning an execution environment.

      • type: "none"

        The type of the object. Always none.

        • "none"

    • OpenAIHosted object { id, capability_directories, desktop, 7 more }

      An environment hosted by OpenAI.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • desktop: object { enabled }

        The effective desktop configuration.

        • enabled: boolean

          Whether the environment provisions a desktop and browser proxy.

      • files: array of HostedEnvironmentFile

        Files available in the environment, excluding their contents.

        • HostedEnvironmentFileID object { id, file_id, path, 2 more }

          A file copied from the OpenAI Files API.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • file_id: string

            The ID of the uploaded file.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "file_id"

            The type of the object. Always file_id.

            • "file_id"

        • Inline object { id, path, size_bytes, type }

          A file supplied inline when the session was created.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • network: object { access, allowed_domains }

        The effective network access policy for the environment.

        • access: "enabled" or "disabled" or "restricted"

          The environment's network access mode.

          • "enabled"

            Allows unrestricted network access.

          • "disabled"

            Disables network access.

          • "restricted"

            Applies the configured domain restrictions.

        • allowed_domains: array of string

          Domains the environment may access when network access is restricted.

      • packages: object { npm, python, system }

        Packages installed in the environment.

        • npm: array of string

          npm packages installed globally in the environment.

        • python: array of string

          Python packages installed in the environment.

        • system: array of string

          System packages installed in the environment.

      • plugins: array of HostedPlugin

        Plugins installed in the environment, excluding their archive contents.

        • description: string

          The installed plugin description.

        • name: string

          The installed plugin name.

        • type: "inline"

          The type of the object. Always inline.

          • "inline"

      • skills: array of HostedSkill

        Skills installed in the environment, excluding their archive contents.

        • HostedSkillReference object { description, name, skill_id, 2 more }

          A skill installed from the Skills API.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • skill_id: string

            The referenced skill ID.

          • type: "skill_reference"

            The type of the object. Always skill_reference.

            • "skill_reference"

          • version: string

            The concrete skill version installed for this session.

        • Inline object { description, name, type }

          A skill installed from an inline ZIP archive.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • type: "openai_hosted"

        The type of the object. Always openai_hosted.

        • "openai_hosted"

      • container_size: optional "small" or "medium" or "large" or null

        The effective CPU and memory tier, or null when unknown or outside the public tiers.

        • "small"

        • "medium"

        • "large"

    • SelfHosted object { id, capability_directories, remote_url, 2 more }

      An environment hosted by the application.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • remote_url: string

        Pass this URL unchanged to codex exec-server --remote when connecting this environment.

      • type: "self_hosted"

        The type of the object. Always self_hosted.

        • "self_hosted"

      • workspace_directory: string

        The absolute project directory inside the environment. Defaults to /workspace.

  • error: string or null

    The error that caused the session to fail, if any.

  • last_active_at: number

    The Unix timestamp, in seconds, when the session was last active.

  • metadata: map[string]

    Custom string key-value pairs attached to the session.

  • object: "agent.session"

    The object type. Always agent.session.

    • "agent.session"

  • required_actions: array of object { request, request_id, turn_id, type } or object { arguments, call_id, name, 2 more } or object { environment_id, type }

    Actions that must be completed before the session can continue.

    • ComputerUseApprovalRequest object { request, request_id, turn_id, type }

      Respond to a computer-use request.

      • request: object { credential_origin, fields, options, 2 more } or object { origin, reason, type }

        The information needed to render the request.

        • BrowserAuthentication object { credential_origin, fields, options, 2 more }

          A registered form awaiting the application's response.

          • credential_origin: string or null

            The registered form or frame origin where values will be entered.

          • fields: array of object { id, label, required, type }

            Controls to render. All submitted values are sensitive.

            • id: string

              The field ID to submit as field_id in a fields entry.

            • label: string

              The label to display beside the control.

            • required: boolean

              Whether this control requires a nonempty value.

            • type: string

              The rendering type, such as email, password, or text.

          • options: array of object { id, field_ids, label }

            Sign-in methods. Empty for a plain form.

            • id: string

              The option ID to submit as selected_option.

            • field_ids: array of string

              IDs from the registered fields that this method accepts.

            • label: string

              The method label to display.

          • reason: string or null

            Why the agent needs the user to sign in.

          • type: "browser_authentication"

            The type of the object. Always browser_authentication.

            • "browser_authentication"

        • BrowserOriginAccess object { origin, reason, type }

          A browser origin awaiting the application's approval decision.

          • origin: string

            The origin the browser needs permission to access.

          • reason: string or null

            The browser's explanation for this request, or null when unavailable.

          • type: "browser_origin_access"

            The type of the object. Always browser_origin_access.

            • "browser_origin_access"

      • request_id: string

        The registered request ID to echo when responding.

      • turn_id: string

        The turn that requested approval.

      • type: "computer_use_approval_request"

        The type of the object. Always computer_use_approval_request.

        • "computer_use_approval_request"

    • FunctionCall object { arguments, call_id, name, 2 more }

      Run a function tool and submit its result.

      • arguments: unknown

        The arguments supplied by the model.

      • call_id: string

        The ID to include when submitting the function result.

      • name: string

        The function name.

      • turn_id: string

        The ID of the turn that requested the function call.

      • type: "function_call"

        The type of the object. Always function_call.

        • "function_call"

    • EnvironmentConnection object { environment_id, type }

      Reconnect a session environment.

      • environment_id: string

        The ID of the environment to reconnect.

      • type: "environment_connection"

        The type of the object. Always environment_connection.

        • "environment_connection"

  • status: "idle" or "in_progress" or "requires_action" or "failed"

    The current status of the session.

    • "idle"

      The session has no turn in progress and is ready for input. A hosted environment may still be provisioning.

    • "in_progress"

      The session is processing a turn.

    • "requires_action"

      The session is waiting for one or more required actions.

    • "failed"

      The session failed.

  • usage: TokenUsage or null

    Best-effort token usage for the session, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

  • vault_ids: array of string

    The IDs of vaults made available to the session.

• type: "agent.session.idle"

  The type of the object. Always agent.session.idle.

  • "agent.session.idle"

Example

{
  "type": "agent.session.idle",
  "event_id": "event_id",
  "session": {
    "metadata": {
      "foo": "string"
    },
    "id": "id",
    "object": "agent.session",
    "created_at": 0,
    "last_active_at": 0,
    "status": "idle",
    "required_actions": [
      {
        "type": "computer_use_approval_request",
        "turn_id": "turn_id",
        "request_id": "request_id",
        "request": {
          "type": "browser_authentication",
          "reason": "reason",
          "credential_origin": "credential_origin",
          "fields": [
            {
              "id": "id",
              "label": "label",
              "type": "type",
              "required": true
            }
          ],
          "options": [
            {
              "id": "id",
              "label": "label",
              "field_ids": [
                "string"
              ]
            }
          ]
        }
      }
    ],
    "error": "error",
    "agent": {
      "id": "id",
      "name": "name",
      "model": "model",
      "reasoning": {
        "effort": "none",
        "summary": "concise"
      },
      "text": {
        "format": {
          "type": "text"
        },
        "verbosity": "low"
      },
      "service_tier": "auto",
      "instructions": "instructions",
      "tools": [
        {
          "type": "function",
          "name": "name",
          "description": "description",
          "parameters": {
            "foo": "bar"
          },
          "defer_loading": true
        }
      ],
      "multi_agent": {
        "enabled": true,
        "max_concurrent_subagents": 1
      }
    },
    "environment": {
      "type": "none"
    },
    "vault_ids": [
      "string"
    ],
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  }
}

agent.session.in_progress

Emitted when a session starts processing a turn.

Schema

Schema name: SessionEventAgentSessionInProgress

• event_id: string

  The unique ID of the event.

• session: AgentSession

  The session that started processing.

  • id: string

    The ID of the session.

  • agent: object { id, instructions, model, 6 more }

    The agent running in the session.

    • id: string

      The ID of the agent.

    • instructions: string or null

      Custom instructions appended to the agent's default base instructions.

    • model: string

      The model used by the agent.

    • multi_agent: MultiAgentConfig

      Configuration for creating and coordinating subagents.

      • enabled: boolean

        Whether subagent tools are enabled. Defaults to false.

      • max_concurrent_subagents: number or null

        Maximum number of subagents that may run concurrently, or null when disabled. Defaults to 6 when enabled.

    • name: string or null

      The reusable agent's name when the session was created, or null if no name was saved. Later changes to the agent's name do not affect this value.

    • reasoning: AgentReasoning

      The agent's reasoning configuration.

      • effort: "none" or "minimal" or "low" or 4 more or null

        The requested reasoning effort, or null when the model selects its own default.

        • "none"

        • "minimal"

        • "low"

        • "medium"

        • "high"

        • "xhigh"

        • "max"

      • summary: "concise" or "detailed" or "auto" or null

        The requested reasoning summary format, or null when summaries are disabled.

        • "concise"

          Returns a concise reasoning summary when supported.

        • "detailed"

          Returns a detailed reasoning summary when supported.

        • "auto"

          Automatically selects the most detailed summary supported by the model.

    • service_tier: "auto" or "default" or "flex" or 3 more

      The effective service-tier policy for model requests. Defaults to auto.

      • "auto"

      • "default"

      • "flex"

      • "priority"

      • "fast"

      • "ultrafast"

    • text: AgentText

      Configuration for text generated by the agent.

      • format: TextFormat

        The effective output format. Defaults to ordinary text.

        • Text object { type }

          Generates ordinary text without a structured-output constraint.

          • type: "text"

            The type of the object. Always text.

            • "text"

        • JSONSchema object { schema, type }

          Constrains generated text to a JSON Schema.

          • schema: map[unknown]

            The JSON Schema that generated text must match.

          • type: "json_schema"

            The type of the object. Always json_schema.

            • "json_schema"

      • verbosity: "low" or "medium" or "high"

        The amount of text produced by the agent. Defaults to medium.

        • "low"

        • "medium"

        • "high"

    • tools: array of AgentTool

      Tools available to the agent.

      • Function object { defer_loading, description, name, 2 more }

        A function defined by the application.

        • defer_loading: boolean

          Whether the function is deferred and discovered through tool search.

        • description: string

          A description of what the function does.

        • name: string

          The name of the function.

        • parameters: map[unknown]

          A JSON Schema object describing the function's arguments.

        • type: "function"

          The type of the object. Always function.

          • "function"

      • ProgrammaticToolCalling object { enabled, type }

        Enables calling tools from model-generated code.

        • enabled: boolean

          Whether tools can be called from model-generated code.

        • type: "programmatic_tool_calling"

          The type of the object. Always programmatic_tool_calling.

          • "programmatic_tool_calling"

      • Mcp object { allowed_tools, connection_origin, credential_id, 5 more }

        Tools provided by a remote MCP server.

        • allowed_tools: array of string or null

          The MCP tools the agent may call.

        • connection_origin: "service" or "environment"

          Where outbound MCP HTTP connections originate.

          • "service"

          • "environment"

        • credential_id: string or null

          The attached vault credential selected for this MCP server, if any. Optional when exactly one attached credential matches the server URL.

        • request_metadata: map[unknown]

          Metadata included with requests to this MCP server.

        • required: boolean

          Whether this MCP server must initialize before the first turn.

        • server_label: string

          A label used to identify the MCP server in tool calls.

        • transport: McpTransport

          The transport used to connect to the MCP server.

          • HTTP object { server_url, type }

            Connects to an MCP server over HTTP.

            • server_url: string

              The URL of the MCP server.

            • type: "http"

              The type of the object. Always http.

              • "http"

          • Stdio object { args, command, cwd, 2 more }

            Starts an MCP server as a local process.

            • args: array of string

              Arguments passed to the MCP server command.

            • command: string

              The command used to start the MCP server.

            • cwd: string

              The working directory used to start the MCP server.

            • env_vars: array of string

              Environment variable names inherited from the execution environment.

            • type: "stdio"

              The type of the object. Always stdio.

              • "stdio"

        • type: "mcp"

          The type of the object. Always mcp.

          • "mcp"

      • WebSearch object { allowed_domains, context_size, location, 2 more }

        Web search.

        • allowed_domains: array of string or null

          Allowed search domains, or null when the search is unrestricted.

        • context_size: "low" or "medium" or "high"

          The amount of search context made available to the model. Defaults to medium.

          • "low"

          • "medium"

          • "high"

        • location: object { city, country, region, timezone } or null

          Approximate location used to localize search results, if provided.

          • city: string or null

            The city name.

          • country: string or null

            The two-letter ISO country code, such as US.

          • region: string or null

            The region or state name.

          • timezone: string or null

            The IANA timezone, such as America/Los_Angeles.

        • mode: "disabled" or "cached" or "live"

          The source used for web search results.

          • "disabled"

          • "cached"

          • "live"

        • type: "web_search"

          The type of the object. Always web_search.

          • "web_search"

      • ComputerUse object { include_screenshots, type }

        Browser use in an OpenAI-hosted session.

        • include_screenshots: boolean

          Whether computer tool outputs include screenshots.

        • type: "computer_use"

          The type of the object. Always computer_use.

          • "computer_use"

  • created_at: number

    The Unix timestamp, in seconds, when the session was created.

  • environment: Environment

    The execution environment for the session.

    • None object { type }

      The session talks to CCA without selecting or provisioning an execution environment.

      • type: "none"

        The type of the object. Always none.

        • "none"

    • OpenAIHosted object { id, capability_directories, desktop, 7 more }

      An environment hosted by OpenAI.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • desktop: object { enabled }

        The effective desktop configuration.

        • enabled: boolean

          Whether the environment provisions a desktop and browser proxy.

      • files: array of HostedEnvironmentFile

        Files available in the environment, excluding their contents.

        • HostedEnvironmentFileID object { id, file_id, path, 2 more }

          A file copied from the OpenAI Files API.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • file_id: string

            The ID of the uploaded file.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "file_id"

            The type of the object. Always file_id.

            • "file_id"

        • Inline object { id, path, size_bytes, type }

          A file supplied inline when the session was created.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • network: object { access, allowed_domains }

        The effective network access policy for the environment.

        • access: "enabled" or "disabled" or "restricted"

          The environment's network access mode.

          • "enabled"

            Allows unrestricted network access.

          • "disabled"

            Disables network access.

          • "restricted"

            Applies the configured domain restrictions.

        • allowed_domains: array of string

          Domains the environment may access when network access is restricted.

      • packages: object { npm, python, system }

        Packages installed in the environment.

        • npm: array of string

          npm packages installed globally in the environment.

        • python: array of string

          Python packages installed in the environment.

        • system: array of string

          System packages installed in the environment.

      • plugins: array of HostedPlugin

        Plugins installed in the environment, excluding their archive contents.

        • description: string

          The installed plugin description.

        • name: string

          The installed plugin name.

        • type: "inline"

          The type of the object. Always inline.

          • "inline"

      • skills: array of HostedSkill

        Skills installed in the environment, excluding their archive contents.

        • HostedSkillReference object { description, name, skill_id, 2 more }

          A skill installed from the Skills API.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • skill_id: string

            The referenced skill ID.

          • type: "skill_reference"

            The type of the object. Always skill_reference.

            • "skill_reference"

          • version: string

            The concrete skill version installed for this session.

        • Inline object { description, name, type }

          A skill installed from an inline ZIP archive.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • type: "openai_hosted"

        The type of the object. Always openai_hosted.

        • "openai_hosted"

      • container_size: optional "small" or "medium" or "large" or null

        The effective CPU and memory tier, or null when unknown or outside the public tiers.

        • "small"

        • "medium"

        • "large"

    • SelfHosted object { id, capability_directories, remote_url, 2 more }

      An environment hosted by the application.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • remote_url: string

        Pass this URL unchanged to codex exec-server --remote when connecting this environment.

      • type: "self_hosted"

        The type of the object. Always self_hosted.

        • "self_hosted"

      • workspace_directory: string

        The absolute project directory inside the environment. Defaults to /workspace.

  • error: string or null

    The error that caused the session to fail, if any.

  • last_active_at: number

    The Unix timestamp, in seconds, when the session was last active.

  • metadata: map[string]

    Custom string key-value pairs attached to the session.

  • object: "agent.session"

    The object type. Always agent.session.

    • "agent.session"

  • required_actions: array of object { request, request_id, turn_id, type } or object { arguments, call_id, name, 2 more } or object { environment_id, type }

    Actions that must be completed before the session can continue.

    • ComputerUseApprovalRequest object { request, request_id, turn_id, type }

      Respond to a computer-use request.

      • request: object { credential_origin, fields, options, 2 more } or object { origin, reason, type }

        The information needed to render the request.

        • BrowserAuthentication object { credential_origin, fields, options, 2 more }

          A registered form awaiting the application's response.

          • credential_origin: string or null

            The registered form or frame origin where values will be entered.

          • fields: array of object { id, label, required, type }

            Controls to render. All submitted values are sensitive.

            • id: string

              The field ID to submit as field_id in a fields entry.

            • label: string

              The label to display beside the control.

            • required: boolean

              Whether this control requires a nonempty value.

            • type: string

              The rendering type, such as email, password, or text.

          • options: array of object { id, field_ids, label }

            Sign-in methods. Empty for a plain form.

            • id: string

              The option ID to submit as selected_option.

            • field_ids: array of string

              IDs from the registered fields that this method accepts.

            • label: string

              The method label to display.

          • reason: string or null

            Why the agent needs the user to sign in.

          • type: "browser_authentication"

            The type of the object. Always browser_authentication.

            • "browser_authentication"

        • BrowserOriginAccess object { origin, reason, type }

          A browser origin awaiting the application's approval decision.

          • origin: string

            The origin the browser needs permission to access.

          • reason: string or null

            The browser's explanation for this request, or null when unavailable.

          • type: "browser_origin_access"

            The type of the object. Always browser_origin_access.

            • "browser_origin_access"

      • request_id: string

        The registered request ID to echo when responding.

      • turn_id: string

        The turn that requested approval.

      • type: "computer_use_approval_request"

        The type of the object. Always computer_use_approval_request.

        • "computer_use_approval_request"

    • FunctionCall object { arguments, call_id, name, 2 more }

      Run a function tool and submit its result.

      • arguments: unknown

        The arguments supplied by the model.

      • call_id: string

        The ID to include when submitting the function result.

      • name: string

        The function name.

      • turn_id: string

        The ID of the turn that requested the function call.

      • type: "function_call"

        The type of the object. Always function_call.

        • "function_call"

    • EnvironmentConnection object { environment_id, type }

      Reconnect a session environment.

      • environment_id: string

        The ID of the environment to reconnect.

      • type: "environment_connection"

        The type of the object. Always environment_connection.

        • "environment_connection"

  • status: "idle" or "in_progress" or "requires_action" or "failed"

    The current status of the session.

    • "idle"

      The session has no turn in progress and is ready for input. A hosted environment may still be provisioning.

    • "in_progress"

      The session is processing a turn.

    • "requires_action"

      The session is waiting for one or more required actions.

    • "failed"

      The session failed.

  • usage: TokenUsage or null

    Best-effort token usage for the session, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

  • vault_ids: array of string

    The IDs of vaults made available to the session.

• type: "agent.session.in_progress"

  The type of the object. Always agent.session.in_progress.

  • "agent.session.in_progress"

Example

{
  "type": "agent.session.in_progress",
  "event_id": "event_id",
  "session": {
    "metadata": {
      "foo": "string"
    },
    "id": "id",
    "object": "agent.session",
    "created_at": 0,
    "last_active_at": 0,
    "status": "idle",
    "required_actions": [
      {
        "type": "computer_use_approval_request",
        "turn_id": "turn_id",
        "request_id": "request_id",
        "request": {
          "type": "browser_authentication",
          "reason": "reason",
          "credential_origin": "credential_origin",
          "fields": [
            {
              "id": "id",
              "label": "label",
              "type": "type",
              "required": true
            }
          ],
          "options": [
            {
              "id": "id",
              "label": "label",
              "field_ids": [
                "string"
              ]
            }
          ]
        }
      }
    ],
    "error": "error",
    "agent": {
      "id": "id",
      "name": "name",
      "model": "model",
      "reasoning": {
        "effort": "none",
        "summary": "concise"
      },
      "text": {
        "format": {
          "type": "text"
        },
        "verbosity": "low"
      },
      "service_tier": "auto",
      "instructions": "instructions",
      "tools": [
        {
          "type": "function",
          "name": "name",
          "description": "description",
          "parameters": {
            "foo": "bar"
          },
          "defer_loading": true
        }
      ],
      "multi_agent": {
        "enabled": true,
        "max_concurrent_subagents": 1
      }
    },
    "environment": {
      "type": "none"
    },
    "vault_ids": [
      "string"
    ],
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  }
}

agent.session.requires_action

Emitted when a session is waiting for one or more required actions.

Schema

Schema name: SessionEventAgentSessionRequiresAction

• event_id: string

  The unique ID of the event.

• session: AgentSession

  The session and its current required actions.

  • id: string

    The ID of the session.

  • agent: object { id, instructions, model, 6 more }

    The agent running in the session.

    • id: string

      The ID of the agent.

    • instructions: string or null

      Custom instructions appended to the agent's default base instructions.

    • model: string

      The model used by the agent.

    • multi_agent: MultiAgentConfig

      Configuration for creating and coordinating subagents.

      • enabled: boolean

        Whether subagent tools are enabled. Defaults to false.

      • max_concurrent_subagents: number or null

        Maximum number of subagents that may run concurrently, or null when disabled. Defaults to 6 when enabled.

    • name: string or null

      The reusable agent's name when the session was created, or null if no name was saved. Later changes to the agent's name do not affect this value.

    • reasoning: AgentReasoning

      The agent's reasoning configuration.

      • effort: "none" or "minimal" or "low" or 4 more or null

        The requested reasoning effort, or null when the model selects its own default.

        • "none"

        • "minimal"

        • "low"

        • "medium"

        • "high"

        • "xhigh"

        • "max"

      • summary: "concise" or "detailed" or "auto" or null

        The requested reasoning summary format, or null when summaries are disabled.

        • "concise"

          Returns a concise reasoning summary when supported.

        • "detailed"

          Returns a detailed reasoning summary when supported.

        • "auto"

          Automatically selects the most detailed summary supported by the model.

    • service_tier: "auto" or "default" or "flex" or 3 more

      The effective service-tier policy for model requests. Defaults to auto.

      • "auto"

      • "default"

      • "flex"

      • "priority"

      • "fast"

      • "ultrafast"

    • text: AgentText

      Configuration for text generated by the agent.

      • format: TextFormat

        The effective output format. Defaults to ordinary text.

        • Text object { type }

          Generates ordinary text without a structured-output constraint.

          • type: "text"

            The type of the object. Always text.

            • "text"

        • JSONSchema object { schema, type }

          Constrains generated text to a JSON Schema.

          • schema: map[unknown]

            The JSON Schema that generated text must match.

          • type: "json_schema"

            The type of the object. Always json_schema.

            • "json_schema"

      • verbosity: "low" or "medium" or "high"

        The amount of text produced by the agent. Defaults to medium.

        • "low"

        • "medium"

        • "high"

    • tools: array of AgentTool

      Tools available to the agent.

      • Function object { defer_loading, description, name, 2 more }

        A function defined by the application.

        • defer_loading: boolean

          Whether the function is deferred and discovered through tool search.

        • description: string

          A description of what the function does.

        • name: string

          The name of the function.

        • parameters: map[unknown]

          A JSON Schema object describing the function's arguments.

        • type: "function"

          The type of the object. Always function.

          • "function"

      • ProgrammaticToolCalling object { enabled, type }

        Enables calling tools from model-generated code.

        • enabled: boolean

          Whether tools can be called from model-generated code.

        • type: "programmatic_tool_calling"

          The type of the object. Always programmatic_tool_calling.

          • "programmatic_tool_calling"

      • Mcp object { allowed_tools, connection_origin, credential_id, 5 more }

        Tools provided by a remote MCP server.

        • allowed_tools: array of string or null

          The MCP tools the agent may call.

        • connection_origin: "service" or "environment"

          Where outbound MCP HTTP connections originate.

          • "service"

          • "environment"

        • credential_id: string or null

          The attached vault credential selected for this MCP server, if any. Optional when exactly one attached credential matches the server URL.

        • request_metadata: map[unknown]

          Metadata included with requests to this MCP server.

        • required: boolean

          Whether this MCP server must initialize before the first turn.

        • server_label: string

          A label used to identify the MCP server in tool calls.

        • transport: McpTransport

          The transport used to connect to the MCP server.

          • HTTP object { server_url, type }

            Connects to an MCP server over HTTP.

            • server_url: string

              The URL of the MCP server.

            • type: "http"

              The type of the object. Always http.

              • "http"

          • Stdio object { args, command, cwd, 2 more }

            Starts an MCP server as a local process.

            • args: array of string

              Arguments passed to the MCP server command.

            • command: string

              The command used to start the MCP server.

            • cwd: string

              The working directory used to start the MCP server.

            • env_vars: array of string

              Environment variable names inherited from the execution environment.

            • type: "stdio"

              The type of the object. Always stdio.

              • "stdio"

        • type: "mcp"

          The type of the object. Always mcp.

          • "mcp"

      • WebSearch object { allowed_domains, context_size, location, 2 more }

        Web search.

        • allowed_domains: array of string or null

          Allowed search domains, or null when the search is unrestricted.

        • context_size: "low" or "medium" or "high"

          The amount of search context made available to the model. Defaults to medium.

          • "low"

          • "medium"

          • "high"

        • location: object { city, country, region, timezone } or null

          Approximate location used to localize search results, if provided.

          • city: string or null

            The city name.

          • country: string or null

            The two-letter ISO country code, such as US.

          • region: string or null

            The region or state name.

          • timezone: string or null

            The IANA timezone, such as America/Los_Angeles.

        • mode: "disabled" or "cached" or "live"

          The source used for web search results.

          • "disabled"

          • "cached"

          • "live"

        • type: "web_search"

          The type of the object. Always web_search.

          • "web_search"

      • ComputerUse object { include_screenshots, type }

        Browser use in an OpenAI-hosted session.

        • include_screenshots: boolean

          Whether computer tool outputs include screenshots.

        • type: "computer_use"

          The type of the object. Always computer_use.

          • "computer_use"

  • created_at: number

    The Unix timestamp, in seconds, when the session was created.

  • environment: Environment

    The execution environment for the session.

    • None object { type }

      The session talks to CCA without selecting or provisioning an execution environment.

      • type: "none"

        The type of the object. Always none.

        • "none"

    • OpenAIHosted object { id, capability_directories, desktop, 7 more }

      An environment hosted by OpenAI.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • desktop: object { enabled }

        The effective desktop configuration.

        • enabled: boolean

          Whether the environment provisions a desktop and browser proxy.

      • files: array of HostedEnvironmentFile

        Files available in the environment, excluding their contents.

        • HostedEnvironmentFileID object { id, file_id, path, 2 more }

          A file copied from the OpenAI Files API.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • file_id: string

            The ID of the uploaded file.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "file_id"

            The type of the object. Always file_id.

            • "file_id"

        • Inline object { id, path, size_bytes, type }

          A file supplied inline when the session was created.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • network: object { access, allowed_domains }

        The effective network access policy for the environment.

        • access: "enabled" or "disabled" or "restricted"

          The environment's network access mode.

          • "enabled"

            Allows unrestricted network access.

          • "disabled"

            Disables network access.

          • "restricted"

            Applies the configured domain restrictions.

        • allowed_domains: array of string

          Domains the environment may access when network access is restricted.

      • packages: object { npm, python, system }

        Packages installed in the environment.

        • npm: array of string

          npm packages installed globally in the environment.

        • python: array of string

          Python packages installed in the environment.

        • system: array of string

          System packages installed in the environment.

      • plugins: array of HostedPlugin

        Plugins installed in the environment, excluding their archive contents.

        • description: string

          The installed plugin description.

        • name: string

          The installed plugin name.

        • type: "inline"

          The type of the object. Always inline.

          • "inline"

      • skills: array of HostedSkill

        Skills installed in the environment, excluding their archive contents.

        • HostedSkillReference object { description, name, skill_id, 2 more }

          A skill installed from the Skills API.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • skill_id: string

            The referenced skill ID.

          • type: "skill_reference"

            The type of the object. Always skill_reference.

            • "skill_reference"

          • version: string

            The concrete skill version installed for this session.

        • Inline object { description, name, type }

          A skill installed from an inline ZIP archive.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • type: "openai_hosted"

        The type of the object. Always openai_hosted.

        • "openai_hosted"

      • container_size: optional "small" or "medium" or "large" or null

        The effective CPU and memory tier, or null when unknown or outside the public tiers.

        • "small"

        • "medium"

        • "large"

    • SelfHosted object { id, capability_directories, remote_url, 2 more }

      An environment hosted by the application.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • remote_url: string

        Pass this URL unchanged to codex exec-server --remote when connecting this environment.

      • type: "self_hosted"

        The type of the object. Always self_hosted.

        • "self_hosted"

      • workspace_directory: string

        The absolute project directory inside the environment. Defaults to /workspace.

  • error: string or null

    The error that caused the session to fail, if any.

  • last_active_at: number

    The Unix timestamp, in seconds, when the session was last active.

  • metadata: map[string]

    Custom string key-value pairs attached to the session.

  • object: "agent.session"

    The object type. Always agent.session.

    • "agent.session"

  • required_actions: array of object { request, request_id, turn_id, type } or object { arguments, call_id, name, 2 more } or object { environment_id, type }

    Actions that must be completed before the session can continue.

    • ComputerUseApprovalRequest object { request, request_id, turn_id, type }

      Respond to a computer-use request.

      • request: object { credential_origin, fields, options, 2 more } or object { origin, reason, type }

        The information needed to render the request.

        • BrowserAuthentication object { credential_origin, fields, options, 2 more }

          A registered form awaiting the application's response.

          • credential_origin: string or null

            The registered form or frame origin where values will be entered.

          • fields: array of object { id, label, required, type }

            Controls to render. All submitted values are sensitive.

            • id: string

              The field ID to submit as field_id in a fields entry.

            • label: string

              The label to display beside the control.

            • required: boolean

              Whether this control requires a nonempty value.

            • type: string

              The rendering type, such as email, password, or text.

          • options: array of object { id, field_ids, label }

            Sign-in methods. Empty for a plain form.

            • id: string

              The option ID to submit as selected_option.

            • field_ids: array of string

              IDs from the registered fields that this method accepts.

            • label: string

              The method label to display.

          • reason: string or null

            Why the agent needs the user to sign in.

          • type: "browser_authentication"

            The type of the object. Always browser_authentication.

            • "browser_authentication"

        • BrowserOriginAccess object { origin, reason, type }

          A browser origin awaiting the application's approval decision.

          • origin: string

            The origin the browser needs permission to access.

          • reason: string or null

            The browser's explanation for this request, or null when unavailable.

          • type: "browser_origin_access"

            The type of the object. Always browser_origin_access.

            • "browser_origin_access"

      • request_id: string

        The registered request ID to echo when responding.

      • turn_id: string

        The turn that requested approval.

      • type: "computer_use_approval_request"

        The type of the object. Always computer_use_approval_request.

        • "computer_use_approval_request"

    • FunctionCall object { arguments, call_id, name, 2 more }

      Run a function tool and submit its result.

      • arguments: unknown

        The arguments supplied by the model.

      • call_id: string

        The ID to include when submitting the function result.

      • name: string

        The function name.

      • turn_id: string

        The ID of the turn that requested the function call.

      • type: "function_call"

        The type of the object. Always function_call.

        • "function_call"

    • EnvironmentConnection object { environment_id, type }

      Reconnect a session environment.

      • environment_id: string

        The ID of the environment to reconnect.

      • type: "environment_connection"

        The type of the object. Always environment_connection.

        • "environment_connection"

  • status: "idle" or "in_progress" or "requires_action" or "failed"

    The current status of the session.

    • "idle"

      The session has no turn in progress and is ready for input. A hosted environment may still be provisioning.

    • "in_progress"

      The session is processing a turn.

    • "requires_action"

      The session is waiting for one or more required actions.

    • "failed"

      The session failed.

  • usage: TokenUsage or null

    Best-effort token usage for the session, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

  • vault_ids: array of string

    The IDs of vaults made available to the session.

• type: "agent.session.requires_action"

  The type of the object. Always agent.session.requires_action.

  • "agent.session.requires_action"

Example

{
  "type": "agent.session.requires_action",
  "event_id": "event_id",
  "session": {
    "metadata": {
      "foo": "string"
    },
    "id": "id",
    "object": "agent.session",
    "created_at": 0,
    "last_active_at": 0,
    "status": "idle",
    "required_actions": [
      {
        "type": "computer_use_approval_request",
        "turn_id": "turn_id",
        "request_id": "request_id",
        "request": {
          "type": "browser_authentication",
          "reason": "reason",
          "credential_origin": "credential_origin",
          "fields": [
            {
              "id": "id",
              "label": "label",
              "type": "type",
              "required": true
            }
          ],
          "options": [
            {
              "id": "id",
              "label": "label",
              "field_ids": [
                "string"
              ]
            }
          ]
        }
      }
    ],
    "error": "error",
    "agent": {
      "id": "id",
      "name": "name",
      "model": "model",
      "reasoning": {
        "effort": "none",
        "summary": "concise"
      },
      "text": {
        "format": {
          "type": "text"
        },
        "verbosity": "low"
      },
      "service_tier": "auto",
      "instructions": "instructions",
      "tools": [
        {
          "type": "function",
          "name": "name",
          "description": "description",
          "parameters": {
            "foo": "bar"
          },
          "defer_loading": true
        }
      ],
      "multi_agent": {
        "enabled": true,
        "max_concurrent_subagents": 1
      }
    },
    "environment": {
      "type": "none"
    },
    "vault_ids": [
      "string"
    ],
    "usage": {
      "input_tokens": 0,
      "input_tokens_details": {
        "cached_tokens": 0
      },
      "output_tokens": 0,
      "output_tokens_details": {
        "reasoning_tokens": 0
      },
      "total_tokens": 0
    }
  }
}

agent.session.failed

Emitted when a session fails.

Schema

Schema name: SessionEventAgentSessionFailed

• event_id: string

  The unique ID of the event.

• session: AgentSession

  The failed session.

  • id: string

    The ID of the session.

  • agent: object { id, instructions, model, 6 more }

    The agent running in the session.

    • id: string

      The ID of the agent.

    • instructions: string or null

      Custom instructions appended to the agent's default base instructions.

    • model: string

      The model used by the agent.

    • multi_agent: MultiAgentConfig

      Configuration for creating and coordinating subagents.

      • enabled: boolean

        Whether subagent tools are enabled. Defaults to false.

      • max_concurrent_subagents: number or null

        Maximum number of subagents that may run concurrently, or null when disabled. Defaults to 6 when enabled.

    • name: string or null

      The reusable agent's name when the session was created, or null if no name was saved. Later changes to the agent's name do not affect this value.

    • reasoning: AgentReasoning

      The agent's reasoning configuration.

      • effort: "none" or "minimal" or "low" or 4 more or null

        The requested reasoning effort, or null when the model selects its own default.

        • "none"

        • "minimal"

        • "low"

        • "medium"

        • "high"

        • "xhigh"

        • "max"

      • summary: "concise" or "detailed" or "auto" or null

        The requested reasoning summary format, or null when summaries are disabled.

        • "concise"

          Returns a concise reasoning summary when supported.

        • "detailed"

          Returns a detailed reasoning summary when supported.

        • "auto"

          Automatically selects the most detailed summary supported by the model.

    • service_tier: "auto" or "default" or "flex" or 3 more

      The effective service-tier policy for model requests. Defaults to auto.

      • "auto"

      • "default"

      • "flex"

      • "priority"

      • "fast"

      • "ultrafast"

    • text: AgentText

      Configuration for text generated by the agent.

      • format: TextFormat

        The effective output format. Defaults to ordinary text.

        • Text object { type }

          Generates ordinary text without a structured-output constraint.

          • type: "text"

            The type of the object. Always text.

            • "text"

        • JSONSchema object { schema, type }

          Constrains generated text to a JSON Schema.

          • schema: map[unknown]

            The JSON Schema that generated text must match.

          • type: "json_schema"

            The type of the object. Always json_schema.

            • "json_schema"

      • verbosity: "low" or "medium" or "high"

        The amount of text produced by the agent. Defaults to medium.

        • "low"

        • "medium"

        • "high"

    • tools: array of AgentTool

      Tools available to the agent.

      • Function object { defer_loading, description, name, 2 more }

        A function defined by the application.

        • defer_loading: boolean

          Whether the function is deferred and discovered through tool search.

        • description: string

          A description of what the function does.

        • name: string

          The name of the function.

        • parameters: map[unknown]

          A JSON Schema object describing the function's arguments.

        • type: "function"

          The type of the object. Always function.

          • "function"

      • ProgrammaticToolCalling object { enabled, type }

        Enables calling tools from model-generated code.

        • enabled: boolean

          Whether tools can be called from model-generated code.

        • type: "programmatic_tool_calling"

          The type of the object. Always programmatic_tool_calling.

          • "programmatic_tool_calling"

      • Mcp object { allowed_tools, connection_origin, credential_id, 5 more }

        Tools provided by a remote MCP server.

        • allowed_tools: array of string or null

          The MCP tools the agent may call.

        • connection_origin: "service" or "environment"

          Where outbound MCP HTTP connections originate.

          • "service"

          • "environment"

        • credential_id: string or null

          The attached vault credential selected for this MCP server, if any. Optional when exactly one attached credential matches the server URL.

        • request_metadata: map[unknown]

          Metadata included with requests to this MCP server.

        • required: boolean

          Whether this MCP server must initialize before the first turn.

        • server_label: string

          A label used to identify the MCP server in tool calls.

        • transport: McpTransport

          The transport used to connect to the MCP server.

          • HTTP object { server_url, type }

            Connects to an MCP server over HTTP.

            • server_url: string

              The URL of the MCP server.

            • type: "http"

              The type of the object. Always http.

              • "http"

          • Stdio object { args, command, cwd, 2 more }

            Starts an MCP server as a local process.

            • args: array of string

              Arguments passed to the MCP server command.

            • command: string

              The command used to start the MCP server.

            • cwd: string

              The working directory used to start the MCP server.

            • env_vars: array of string

              Environment variable names inherited from the execution environment.

            • type: "stdio"

              The type of the object. Always stdio.

              • "stdio"

        • type: "mcp"

          The type of the object. Always mcp.

          • "mcp"

      • WebSearch object { allowed_domains, context_size, location, 2 more }

        Web search.

        • allowed_domains: array of string or null

          Allowed search domains, or null when the search is unrestricted.

        • context_size: "low" or "medium" or "high"

          The amount of search context made available to the model. Defaults to medium.

          • "low"

          • "medium"

          • "high"

        • location: object { city, country, region, timezone } or null

          Approximate location used to localize search results, if provided.

          • city: string or null

            The city name.

          • country: string or null

            The two-letter ISO country code, such as US.

          • region: string or null

            The region or state name.

          • timezone: string or null

            The IANA timezone, such as America/Los_Angeles.

        • mode: "disabled" or "cached" or "live"

          The source used for web search results.

          • "disabled"

          • "cached"

          • "live"

        • type: "web_search"

          The type of the object. Always web_search.

          • "web_search"

      • ComputerUse object { include_screenshots, type }

        Browser use in an OpenAI-hosted session.

        • include_screenshots: boolean

          Whether computer tool outputs include screenshots.

        • type: "computer_use"

          The type of the object. Always computer_use.

          • "computer_use"

  • created_at: number

    The Unix timestamp, in seconds, when the session was created.

  • environment: Environment

    The execution environment for the session.

    • None object { type }

      The session talks to CCA without selecting or provisioning an execution environment.

      • type: "none"

        The type of the object. Always none.

        • "none"

    • OpenAIHosted object { id, capability_directories, desktop, 7 more }

      An environment hosted by OpenAI.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • desktop: object { enabled }

        The effective desktop configuration.

        • enabled: boolean

          Whether the environment provisions a desktop and browser proxy.

      • files: array of HostedEnvironmentFile

        Files available in the environment, excluding their contents.

        • HostedEnvironmentFileID object { id, file_id, path, 2 more }

          A file copied from the OpenAI Files API.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • file_id: string

            The ID of the uploaded file.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "file_id"

            The type of the object. Always file_id.

            • "file_id"

        • Inline object { id, path, size_bytes, type }

          A file supplied inline when the session was created.

          • id: string

            The session-scoped ID of the file in the execution environment.

          • path: string

            The file's absolute path inside the environment.

          • size_bytes: number

            The decoded file size in bytes.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • network: object { access, allowed_domains }

        The effective network access policy for the environment.

        • access: "enabled" or "disabled" or "restricted"

          The environment's network access mode.

          • "enabled"

            Allows unrestricted network access.

          • "disabled"

            Disables network access.

          • "restricted"

            Applies the configured domain restrictions.

        • allowed_domains: array of string

          Domains the environment may access when network access is restricted.

      • packages: object { npm, python, system }

        Packages installed in the environment.

        • npm: array of string

          npm packages installed globally in the environment.

        • python: array of string

          Python packages installed in the environment.

        • system: array of string

          System packages installed in the environment.

      • plugins: array of HostedPlugin

        Plugins installed in the environment, excluding their archive contents.

        • description: string

          The installed plugin description.

        • name: string

          The installed plugin name.

        • type: "inline"

          The type of the object. Always inline.

          • "inline"

      • skills: array of HostedSkill

        Skills installed in the environment, excluding their archive contents.

        • HostedSkillReference object { description, name, skill_id, 2 more }

          A skill installed from the Skills API.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • skill_id: string

            The referenced skill ID.

          • type: "skill_reference"

            The type of the object. Always skill_reference.

            • "skill_reference"

          • version: string

            The concrete skill version installed for this session.

        • Inline object { description, name, type }

          A skill installed from an inline ZIP archive.

          • description: string

            The installed skill description.

          • name: string

            The installed skill name.

          • type: "inline"

            The type of the object. Always inline.

            • "inline"

      • type: "openai_hosted"

        The type of the object. Always openai_hosted.

        • "openai_hosted"

      • container_size: optional "small" or "medium" or "large" or null

        The effective CPU and memory tier, or null when unknown or outside the public tiers.

        • "small"

        • "medium"

        • "large"

    • SelfHosted object { id, capability_directories, remote_url, 2 more }

      An environment hosted by the application.

      • id: string

        The public ID of the environment.

      • capability_directories: array of string

        Directories that contain capabilities exposed to the agent.

      • remote_url: string

        Pass this URL unchanged to codex exec-server --remote when connecting this environment.

      • type: "self_hosted"

        The type of the object. Always self_hosted.

        • "self_hosted"

      • workspace_directory: string

        The absolute project directory inside the environment. Defaults to /workspace.

  • error: string or null

    The error that caused the session to fail, if any.

  • last_active_at: number

    The Unix timestamp, in seconds, when the session was last active.

  • metadata: map[string]

    Custom string key-value pairs attached to the session.

  • object: "agent.session"

    The object type. Always agent.session.

    • "agent.session"

  • required_actions: array of object { request, request_id, turn_id, type } or object { arguments, call_id, name, 2 more } or object { environment_id, type }

    Actions that must be completed before the session can continue.

    • ComputerUseApprovalRequest object { request, request_id, turn_id, type }

      Respond to a computer-use request.

      • request: object { credential_origin, fields, options, 2 more } or object { origin, reason, type }

        The information needed to render the request.

        • BrowserAuthentication object { credential_origin, fields, options, 2 more }

          A registered form awaiting the application's response.

          • credential_origin: string or null

            The registered form or frame origin where values will be entered.

          • fields: array of object { id, label, required, type }

            Controls to render. All submitted values are sensitive.

            • id: string

              The field ID to submit as field_id in a fields entry.

            • label: string

              The label to display beside the control.

            • required: boolean

              Whether this control requires a nonempty value.

            • type: string

              The rendering type, such as email, password, or text.

          • options: array of object { id, field_ids, label }

            Sign-in methods. Empty for a plain form.

            • id: string

              The option ID to submit as selected_option.

            • field_ids: array of string

              IDs from the registered fields that this method accepts.

            • label: string

              The method label to display.

          • reason: string or null

            Why the agent needs the user to sign in.

          • type: "browser_authentication"

            The type of the object. Always browser_authentication.

            • "browser_authentication"

        • BrowserOriginAccess object { origin, reason, type }

          A browser origin awaiting the application's approval decision.

          • origin: string

            The origin the browser needs permission to access.

          • reason: string or null

            The browser's explanation for this request, or null when unavailable.

          • type: "browser_origin_access"

            The type of the object. Always browser_origin_access.

            • "browser_origin_access"

      • request_id: string

        The registered request ID to echo when responding.

      • turn_id: string

        The turn that requested approval.

      • type: "computer_use_approval_request"

        The type of the object. Always computer_use_approval_request.

        • "computer_use_approval_request"

    • FunctionCall object { arguments, call_id, name, 2 more }

      Run a function tool and submit its result.

      • arguments: unknown

        The arguments supplied by the model.

      • call_id: string

        The ID to include when submitting the function result.

      • name: string

        The function name.

      • turn_id: string

        The ID of the turn that requested the function call.

      • type: "function_call"

        The type of the object. Always function_call.

        • "function_call"

    • EnvironmentConnection object { environment_id, type }

      Reconnect a session environment.

      • environment_id: string

        The ID of the environment to reconnect.

      • type: "environment_connection"

        The type of the object. Always environment_connection.

        • "environment_connection"

  • status: "idle" or "in_progress" or "requires_action" or "failed"

    The current status of the session.

    • "idle"

      The session has no turn in progress and is ready for input. A hosted environment may still be provisioning.

    • "in_progress"

      The session is processing a turn.

    • "requires_action"

      The session is waiting for one or more required actions.

    • "failed"

      The session failed.

  • usage: TokenUsage or null

    Best-effort token usage for the session, or null if unknown. Recorded usage may change.

    • input_tokens: number

      The number of input tokens used by the agent.

    • input_tokens_details: object { cached_tokens }

      A breakdown of the agent's input token usage.

      • cached_tokens: number

        The number of input tokens retrieved from the prompt cache.

    • output_tokens: number

      The number of output tokens generated by the agent.

    • output_tokens_details: object { reasoning_tokens }

      A breakdown of the agent's output token usage.

      • reasoning_tokens: number

        The number of output tokens used for reasoning.

    • total_tokens: number

      The total number of input and output tokens used by the agent.

  • vault_ids: array of string

    The IDs of vaults made available to the session.

• type: "agent.session.failed"

  The type of the object. Always agent.session.failed.

  • "agent.session.failed"