GenBatch provides a hosted remote MCP server for supported clients. It lets an AI assistant submit image or video work, check a job later, and read the available credit balance without requiring you to run a local generation server.
What the connection adds
After you connect GenBatch, your MCP client can discover the current tool schema directly. The tool set covers image generation, text-to-video, image-to-video, job polling, and credit-balance reads. It uses the same GenBatch account and balance as the web app.
The connected tool schema is the source of truth for its accepted arguments. That matters because an assistant can inspect the schema before it calls a tool instead of guessing model names, aspect-ratio values, or batch limits.
Connect in three steps
- Open Connect while signed in to GenBatch.
- Use the sign-in flow when your MCP client supports it. Otherwise, create a revocable GenBatch access key and follow the manual configuration shown for your client.
- Ask the client to list the GenBatch tools, then start with one inexpensive test request.
Treat a manually created key like a password. Do not paste it into a prompt, commit it to a repository, or place it in client-side code. Revoke a key from the Connect page if it is exposed or no longer used.
A reliable agent workflow
Generation is asynchronous, so a good assistant should submit a job, keep its ID, and poll that job until it reaches a final state. It should not repeatedly submit the same request merely because an image or video is not ready yet.
For a larger batch, ask the assistant to:
- confirm the number of requested outputs before submission;
- preserve stable IDs for each prompt or scene;
- report the expected credit requirement;
- return the job ID immediately;
- poll with a sensible delay; and
- summarize failed items separately from completed results.
MCP or REST API?
Use MCP when a person is directing work through an assistant. Use the REST API when your application owns the loop, needs predictable JSON handling, or schedules work without a conversation. Both paths use the same account and credits.
GenBatch charges credits after successful generation and does not require a separate model-provider API key. See the detailed API and MCP reference for current limits and pricing for current costs.
