Midjourney MCP

What midjourney-mcp MCP server does

The midjourney-mcp MCP server gives an MCP-compatible AI client access to Midjourney operations through RunAPI. It is scoped to the Midjourney model line instead of exposing RunAPI's broader model catalog. The available operations cover image generation, image editing, image-to-video conversion, video extension, prompt processing, seed retrieval, task status, and pricing.

Generation tools can return immediately with a task identifier or wait until a terminal state is available. Results from asynchronous task tools include the task ID, current status, and output URLs when available. Synchronous operations return their operation result directly.

How it works

The server runs as a standard stdio MCP process, so a compatible host starts it with npx and communicates with it over the MCP connection. An agent can inspect current pricing, submit a Midjourney task, and then use the returned identifier with get_task if it chose not to wait.

The server supports the model identifiers documented for its operations. Text-to-image uses midjourney-v8.1, image editing uses midjourney-edit-image, and image-to-video plus video extension use midjourney-image-to-video. Model availability may change, so pricing or the current Midjourney model page should be checked before relying on a specific model.

Setup and configuration

Install the package by registering the stdio command with an MCP host:

{
  "mcpServers": {
    "midjourney": {
      "command": "npx",
      "args": ["-y", "@runapi.ai/midjourney-mcp"]
    }
  }
}

Claude Code can register the same package with claude mcp add, using either user or project scope. The README also provides examples for Codex, Cursor, Windsurf, VS Code, and Roo Code.

For interactive authentication, ask the assistant to use the RunAPI login flow. It opens a browser and stores credentials in ~/.config/runapi/config.json, which is also used by runapi login. Headless and CI environments can provide RUNAPI_API_KEY before starting the MCP host. A pre-provisioned configuration file may contain an apiKey value. Do not commit credentials. Pricing checks do not require sign-in.

Tools and capabilities

• text_to_image: Creates a Midjourney text-to-image task and optionally waits for completion.
• edit_image: Creates an image-editing task and optionally waits for a terminal status.
• image_to_video: Creates an image-to-video task and optionally waits.
• extend_video: Extends a Midjourney video and optionally waits.
• get_task: Retrieves the latest status and payload for an existing task.
• get_seed: Runs a synchronous seed lookup.
• image_to_prompt: Runs a synchronous image-to-prompt operation.
• shorten_prompt: Runs a synchronous prompt-shortening operation.
• check_pricing: Looks up current pricing for a Midjourney model and endpoint without authentication.

Limitations and notes

Task creation and status polling require RunAPI authentication; pricing does not. When a create operation is submitted without waiting, the client must retain the returned task ID and call get_task later. The documented model list and availability can change between releases, so applications should avoid assuming that a model or endpoint remains unchanged.