---
title: "Execute agent"
method: POST
path: "/agents/execute"
tags: ["Agents"]
---

# Execute agent

`POST /agents/execute`

Executes an agent with optional content.

## Request body

- object — Request parameters for executing an agent
  - `agentId` string — Optional - the backend will automatically get or create an agent.
  - `queueId` string — Optional queue ID. When provided, the execution is enqueued onto that queue: it is admitted against the queue concurrency limit (starting immediately or waiting in line), prefixed with the queue default message, and runs on the queue Kaizen version. When omitted, the execution starts immediately under the organization concurrency limit. When set and no agentId is given, the agent is resolved from the queue (the agent already running work on it, else the organization most recent agent).
  - `content` union[] — Optional content array. If provided, this will be used as the initial message. Each item must have a `type` field with one of the following values: - `text` — Plain text content. Requires a `text` string field. - `file` — A previously uploaded file. Requires an `id` string field with the file external ID. - `skill` — A skill to inject into the prompt. Requires an `id` string field with the agent_fs_node external ID. - `login` — A login to pin to this execution. Requires an `id` string field with the login ID. The login is validated against the organization and injected into the initial prompt so the agent uses it without searching for credentials. - `json` — Structured key-value data. Requires a `data` object field. Rendered as a table in the UI and persisted as input.json.
    - union
      - object
        - `type` 'text', required
        - `text` string, required — Text content
      - object
        - `type` 'file', required
        - `id` string, required — External file ID referencing a previously uploaded file.
      - object
        - `type` 'skill', required
        - `id` string, required — ID of the skill to use (the agent_fs_node external ID). The skill content will be read and injected into the initial user prompt.
      - object
        - `type` 'login', required
        - `id` string, required — ID of a login to pin to this execution. The login is validated against the organization and injected into the initial prompt so the agent uses it without having to search for credentials.
      - object
        - `type` 'json', required
        - `data` object, required — Structured JSON data as key-value pairs. Rendered as a table in the UI and persisted as input.json in the session folder.
  - `archiveSessionFiles` boolean — When true, request a background archive of the thread session files after the execution reaches a terminal state.
  - `summarySchema` object — Optional JSON Schema for per-execution summary extraction. Overrides the agent-level result schema for this execution only.
  - `kaizenVersion` 'Full' | 'Lite' — Kaizen version tier to use for this execution. Determines the default model based on your organization’s model alias configuration for the selected tier. - `Full` — Uses the model configured for Kaizen Full (typically a frontier model). - `Lite` — Uses the model configured for Kaizen Lite (typically a smaller, faster model). When not provided, the organization’s default Kaizen version is used.
  - `enableComputerUse` boolean — Whether the computer-use tool is available for this execution. When false, computer use is disabled even if the org-level flag is on. When omitted, the org-level flag applies.
  - `networkPolicy` object — Network settings for the execution browser session.
    - `staticIpId` string — ID of a static IP to route this execution through. The execution is pinned to that IP and proxy usage is forced. Must be an active static IP available to your organization.

## Response `200`

Agent execution started successfully

- object — Response after starting agent execution
  - `sessionId` string, required — ID of the session started for this execution
  - `executionId` string, nullable, required — ID of the execution, or null when the work was enqueued onto a queue (a queued execution has no standalone execution record).
  - `status` string, required — Current status of the execution. When enqueued onto a queue, this is InProgress when it started immediately or WaitingInLine when the queue was at capacity.

## Other responses

- `400` — Invalid request parameters
- `404` — The supplied queueId or agentId could not be resolved for the organization, or no agent could be resolved for the queue. Codes: AGENT_QUEUE_NOT_FOUND, QUEUE_AGENT_NOT_FOUND, QUEUE_HAS_NO_RESOLVABLE_AGENT.
- `500` — Internal server error

---

[API](https://skmtc.dev/kaizenautomation/apis/kaizen-api.md) · [All operations](https://skmtc.dev/kaizenautomation/apis/kaizen-api/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/kaizenautomation/kaizen-api/revisions/4aa3664ceff2/schema)