> ## Documentation Index > Fetch the complete documentation index at: https://docs.agentmark.co/llms.txt > Use this file to discover all available pages before exploring further. # Launch a worker run > Launches a terminal coding agent against the app's connected repo on managed compute: it clones the code, does the task, and delivers a branch + pull request. Poll `GET /v1/workers/runs/{runId}` for the outcome, or hand a human the returned `url` — the dashboard's live chat thread for the run. Requires the key's environment (or the app default) to carry the agent's credential env var (e.g. `ANTHROPIC_API_KEY`). ## OpenAPI ````yaml /openapi.yaml post /v1/workers/runs openapi: 3.0.3 info: contact: email: hello@agentmark.co url: https://docs.agentmark.co description: >- The AgentMark Gateway API lets you ingest traces, create scores, and retrieve prompt templates programmatically. Most developers should use the [AgentMark SDK](/introduction/overview) for integration. The REST API is for cases where you need direct HTTP access or are building a custom integration. Versioning: every endpoint is prefixed with `/v1/`. Breaking changes ship under a new version prefix (`/v2/`, etc.) with a 90-day-minimum deprecation window — see [API versioning & stability](/api-reference/versioning) for the full policy on what constitutes a breaking vs. additive change. title: AgentMark Gateway API version: '1.0' servers: - description: Production (AgentMark Cloud) url: https://api.agentmark.co - description: Local dev server (`npx @agentmark-ai/cli dev`) url: http://localhost:9418 security: - AppId: [] BearerAuth: [] tags: - description: Ingest, query, and export OpenTelemetry trace data. name: Traces - description: >- Coding-agent session ingest (outerlayer sync) and content-addressed session images. name: Agents - description: Query individual spans across traces. name: Spans - description: Create, retrieve, list, and delete score records for spans and traces. name: Scoring - description: Structured-filter search across observability resources. name: Search - description: Retrieve the effective synced project configuration. name: Config - description: List and retrieve datasets. name: Datasets - description: Query server feature availability. name: Capabilities - description: Per-model LLM pricing data. Public, unauthenticated. name: Pricing - description: >- Create, list, and revoke tenant API keys. Plaintext returned exactly once at creation. name: API Keys - description: >- List and retrieve score configuration definitions from the synced project config. name: Score Configs - description: >- Trigger and inspect managed deployments. POST returns 202; deployments run asynchronously and progress is observed via GET. name: Deployments - description: >- Per-app named environments (e.g. `dev`, `prod`). CRUD for env lifecycle; promote/rollback routes land alongside in this group as feature 054 lands the saga. name: Environments - description: >- App CRUD — the top-level tenant entity every other resource hangs off. Lets a headless agent provision an app without the dashboard. name: Apps - description: >- Cloud workers — terminal coding agents on managed compute. Launch one-shot runs or persistent multi-turn sessions against the app's connected repo; every response carries the dashboard deep link to the live thread. name: Workers - description: Service health checks. name: Health paths: /v1/workers/runs: post: tags: - Workers summary: Launch a worker run description: >- Launches a terminal coding agent against the app's connected repo on managed compute: it clones the code, does the task, and delivers a branch + pull request. Poll `GET /v1/workers/runs/{runId}` for the outcome, or hand a human the returned `url` — the dashboard's live chat thread for the run. Requires the key's environment (or the app default) to carry the agent's credential env var (e.g. `ANTHROPIC_API_KEY`). operationId: launch-worker-run requestBody: content: application/json: schema: properties: agent: description: >- Agent id (default claude-code). One of: claude-code, codex, cursor, opencode type: string base_branch: description: Base branch (defaults to the app's tracked branch). type: string task_prompt: description: What the agent should do. minLength: 1 type: string wall_clock_cap_s: description: Wall-clock cap in seconds (default 1800). maximum: 3600 minimum: 60 type: integer required: - task_prompt type: object responses: '200': content: application/json: schema: properties: data: properties: run_id: type: string status: type: string url: description: Dashboard deep link to the live run thread. nullable: true type: string required: - run_id - status - url type: object required: - data type: object description: Run dispatched. '400': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Invalid request body or unknown agent. '401': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Missing or invalid API key. '402': content: application/json: schema: properties: error: properties: code: enum: - entitlement_required type: string current: type: number entitlement: type: string limit: type: number message: type: string required: - code - message - entitlement type: object required: - error type: object description: >- Workers not enabled for the tenant's tier, or a concurrency / monthly-minutes quota is exhausted. '403': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Caller lacks 'worker_run.insert' permission. '502': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Dispatch failed; the run is recorded as failed. '503': content: application/json: schema: properties: error: properties: code: type: string message: type: string required: - code - message type: object required: - error type: object description: Worker compute is not configured on this deployment. components: securitySchemes: AppId: description: >- Application ID for tenant scoping. X-Outerlayer-App-Id is accepted as an equivalent header. in: header name: X-Agentmark-App-Id type: apiKey BearerAuth: description: API key (sk_agentmark_*) scheme: bearer type: http ````