EN
Sign inTry for freeSign in

Banner generation API and MCP for AI agents

Everything I do in the web app is available from code: REST API v1 and an MCP server for AI agents such as Claude, ChatGPT and Cursor.

REST API

  • POST /v1/jobs — a task in any mode: banner, resize, platform pack, background, translation, video, HTML5.
  • POST /v1/estimate — the cost in credits before you start.
  • GET /v1/jobs/{id} — status, Neuro’s questions and signed file links; a ZIP at zip_url.
  • Webhooks job.step, job.completed and job.failed, signed with HMAC-SHA256.
  • The Idempotency-Key header prevents duplicate runs; with autostart: true, a task short of credits waits as a draft and starts on its own.

MCP for AI agents

The MCP server speaks Streamable HTTP with an API key: an agent uploads a photo by URL, gets an estimate, starts a task and answers my questions. 22 tools cover every task, files, the balance and brand kits.

Keys and testing

nb_live_ keys spend the workspace’s credits; nb_test_ keys are a stub — the task goes through every check without generating or charging anything. Credits are charged only for files that pass the AI check.

API and MCP documentation

I will build these formats from your photo50 credits on sign-up, no card. Credits are charged only for files that pass the AI check.Try for free

Questions

Is there a sandbox?

Yes: an nb_test_ key answers like a real one but generates and charges nothing.

How do I know a task is done?

Subscribe to the job.completed webhook or poll GET /v1/jobs/{id}.