Connect your AI agent

Use GenQueue from ChatGPT, Claude, Cursor, or your own MCP client with the same account permissions, credits, and publishing safeguards as the web app.

Remote MCP server URL

https://www.genqueue.com/api/mcp

Connect a supported client

  1. 1. ChatGPT

    Open Settings → Apps & Connectors → Advanced settings, enable Developer mode, create an MCP app, and use the server URL below with OAuth.

  2. 2. Claude

    Open Settings → Connectors, add a custom connector, enter the server URL, then complete the GenQueue sign-in and consent flow.

  3. 3. Cursor

    Add a remote MCP server in Cursor settings with the server URL, authenticate in the browser, and use Agent mode when invoking tools.

During consent, select only the capabilities the client needs. Immediate publishing is separately granted and remains disabled unless you explicitly enable it.

How authentication works

OAuth clients open GenQueue in your browser. Sign into the GenQueue account you want the agent to use, review the requested capabilities, and approve access. The connection is bound to that signed-in GenQueue user and account—not to the email address used by the AI client.

You can review, reduce, or revoke a connection under GenQueue Settings → Automation. Revocation takes effect at GenQueue even if a client still holds an unexpired token.

API keys for custom clients

In GenQueue, open Settings → Automation and create a personal API key with the minimum required capabilities. Send it only from a trusted server or native MCP client as Authorization: Bearer YOUR_KEY.

The key is shown once. Never paste it into prompts, browser JavaScript, source control, logs, or screenshots. Revoke it immediately if it may have been exposed.

REST API for scripts and services

REST v1 uses scoped personal API keys created in Settings → Automation. MCP OAuth tokens are resource-bound to the MCP endpoint and are intentionally not accepted by REST. The OpenAPI document is available at https://www.genqueue.com/api/v1/openapi.json.

curl https://www.genqueue.com/api/v1/models?type=image \
  -H "Authorization: Bearer gq_live_YOUR_KEY"
curl -X POST https://www.genqueue.com/api/v1/generations/images \
  -H "Authorization: Bearer gq_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"request_id":"NEW_UUID","prompt":"a studio portrait","model_slug":"gpt-image-2","count":1}'

Retries and polling

Use a new UUID request_id for each new mutation. Retry an uncertain request with the identical UUID and arguments to receive the original result. Poll job and post status endpoints until the asynchronous work reaches a terminal state.

Errors and pagination

Errors return a bounded code and message. A 409 means the request identity or resource state conflicts; a 429 includes retry guidance. Pass returned opaque cursors unchanged when requesting the next page.

Available tools

Tool discovery reflects the capabilities granted to the connection. Mutations use a UUID request_id: generate a new UUID for a new action and reuse the same UUID and identical arguments only when retrying that action.

Account and discovery

  • whoami
  • list_models
  • get_credit_balance

Jobs and media

  • get_job_status
  • list_jobs
  • list_library
  • import_reference

Generation

  • generate_image
  • generate_video
  • generate_caption

Social and scheduling

  • list_social_accounts
  • get_posting_schedule
  • schedule_post
  • update_scheduled_post
  • cancel_scheduled_post
  • list_posts
  • get_post_status
  • publish_post_now
Generation and publishing are asynchronous. Use get_job_status after generation and get_post_status after publishing. Scheduling is the safe default; publish_post_now requires its own grant and confirm: true.

Need help? Contact GenQueue support.