Batch Banana MCP server

The Batch Banana MCP server connects an AI assistant, such as ChatGPT or Cursor, to your Batch Banana workspace. Through it, the assistant can:

Generate images and video with the models in Batch Banana.
Edit images: background removal, upscales, relighting, expansion, and more.
Search your library for existing images and video to build on.
List your projects and file new work under one.

The server speaks the Model Context Protocol. Its address is:

https://mcp.batchbanana.com/mcp

Requirements

A Batch Banana account. You sign in during setup. If you don't have one, create one first.
Credits in the workspace you connect. Generations and paid edits spend that workspace's credits, the same credits you use in the Batch Banana app. Listing models, searching your library, and getting quotes are free.
Welcome credits. A new account gets 100 credits in its default workspace once its email address is verified.

When you run out, see plans and credits.

Connect

In your AI app, add a custom connector or MCP server with the address https://mcp.batchbanana.com/mcp. Clients that use a config file take this entry:
```
{
  "mcpServers": {
    "batch-banana": { "url": "https://mcp.batchbanana.com/mcp" }
  }
}
```
The app opens a Batch Banana page in your browser. Sign in.
The consent screen names the app and the workspace it will get access to. If you belong to more than one workspace, pick one.
Review the permissions and click Allow access. The app returns to your conversation, connected.

Each connection covers one workspace. To use another workspace, connect again and pick it.

There is no API key to paste. The app signs in with OAuth and never sees your password.

Disconnect

Open Batch Banana, go to your workspace's Settings, then MCP. Under Connected apps, click Disconnect next to the app. The app loses access on its next request.

The member who connected an app can disconnect it, and the workspace owner can disconnect any app. Deleting a workspace disconnects every app connected to it.

Tools

Tool	What it does
`list_models`	Lists image and video generation models with pricing and capabilities.
`describe_model`	Returns the full parameter definitions for one model.
`list_edit_operations`	Lists image edit operations (background removal, upscales, relight, expand, and more) and cost.
`describe_edit_operation`	Returns the full parameter definitions for one edit operation.
`quote_image_generation`	Prices an image generation without charging.
`quote_video_generation`	Prices a video generation without charging.
`quote_image_edit`	Prices an image edit without charging.
`list_projects`	Lists the workspace's projects, newest first.
`search_library`	Finds the workspace's existing images and video, by prompt or newest first.
`get_file`	Returns one stored file's URLs, content type, size, and dimensions.
`import_image`	Imports a public https image URL into the workspace and returns a file id.
`generate_image`	Starts an image generation. Spends credits.
`generate_video`	Starts a video generation. Spends credits.
`edit_image`	Runs an edit operation on stored images. Paid operations spend credits.
`get_generation`	Returns a job's status and, once finished, its output URLs.

No tool deletes anything.

How a generation works

Pick a model. list_models, then describe_model for its options.
Gather inputs (optional). search_library finds earlier work to use as a reference. import_image turns an image URL into a file id, since generation and edit tools take file ids, not URLs.
Quote. quote_image_generation, quote_video_generation, or quote_image_edit returns the cost in credits without charging.
Generate. generate_image, generate_video, or edit_image with an idempotencyKey. Credits are charged when the job starts. If the assistant passes the quoted cost as expectedCreditCost and the price changed since the quote, the request is rejected instead of charged.
Wait for the result. get_generation with the job id until the status is completed or failed. Outputs are URLs to the files in your library.

The idempotencyKey makes retries safe. Sending the same request again with the same key returns the original job and does not charge twice.

Everything the assistant creates appears in your Batch Banana library, like work you create in the app.

Errors

Error	What it means
`INSUFFICIENT_CREDITS`	The workspace doesn't have enough credits for the request. Add credits in Batch Banana; see plans and credits.
`RATE_LIMITED`	Too many requests in a short time. Wait a moment and try again.
`PRICE_CHANGED`	The cost changed since the quote. Quote again.
`IDEMPOTENCY_CONFLICT`	The `idempotencyKey` was already used for a different request. Use a new key.
`INVALID_ARGUMENT`	A parameter is missing or not valid for the model. `describe_model` lists the valid options.
`NOT_FOUND`	The model, file, project, or job doesn't exist or belongs to another workspace.
`NOT_AUTHORIZED`	The connection doesn't have permission for the request.
`UNAUTHENTICATED`	The connection was disconnected. Reconnect the app.

A job can also end with status failed, for example when the model provider refuses the prompt or image under its content policy. When a job fails, the credits it used are returned to the workspace automatically.

Data and privacy

A connected app works inside the one workspace you approved. It can:

Read the model and edit operation catalog, including prices.
Read the workspace's projects, its library of images and video with their prompts, file details, and job status.
Start image generations, video generations, and image edits, which spend the workspace's credits.
Import images from public URLs into the workspace.

The permissions behind this are catalog:read, assets:read, assets:write, generations:read, generations:create, and edits:create. Every connection gets all six, because the tools need them.

It cannot read other workspaces you belong to, change your account or billing, or delete content.

The AI app you connect is run by its own provider, which has its own privacy policy. Read our Privacy Policy and Terms of Service for how Batch Banana handles your data.

Support

Email support@batchbanana.com or visit the support page.