Midjourney MCP Server & CLI

What Midjourney MCP Server & CLI does

Midjourney MCP Server & CLI provides two interfaces for working with Midjourney: an MCP server for AI applications and a command-line program for terminals, scripts, scheduled jobs, and pipelines. Both surfaces use the same underlying tool definitions, so the operations available through one interface correspond to commands or MCP tools on the other.

The main workflow is image generation. An agent or script can submit a prompt, wait for rendering to finish, inspect the resulting job, and download the generated files. The server also covers job history, the active queue, reruns, image variations, upscaling or other job actions exposed by Midjourney, and moodboards that can be reused as a style reference.

How it works

Midjourney does not provide an API for this use case, so Midjourney MCP Server & CLI controls a real Chrome browser. The browser uses a separate profile at ~/.midjourney-mcp/chrome-profile; signing in through that profile establishes the session used by later commands and MCP calls. The normal Chrome profile and its browsing data are not used.

The MCP process is normally launched by an MCP client rather than started manually. The CLI runs directly in a terminal and returns structured data. JSON output, field selection, compact formatting, and JSON errors on stderr make the command surface suitable for shell pipelines and tools such as jq.

Generation can either wait for completion or return before the job finishes. A separate wait operation can then monitor the job. Completed results can be written to disk, allowing later pipeline steps to consume image files instead of only receiving a job identifier or browser screenshot.

Setup and configuration

Install and run the one-time login command with npx:

npx -y @thenavidm/midjourney-mcp-cli@latest login

A Chrome window opens on Midjourney. After signing in, the session remains in the dedicated browser profile. The MCP client can then be configured with:

claude mcp add midjourney -- npx -y @thenavidm/midjourney-mcp-cli@latest

No value is passed with -e, and the README states that there is no API key or exported cookie to configure. The CLI also includes doctor for setup checks. To revoke the stored session, sign out in the managed browser or remove the profile directory.

Tools and capabilities

The Midjourney MCP Server & CLI tool set includes operations for:

• Generating images synchronously with imagine or submitting work with submit_imagine.
• Checking identity and session status with whoami.
• Reading individual jobs, recent jobs, and the current queue.
• Waiting for a job, rerunning it, or creating variations from an image.
• Downloading completed results by job or URL.
• Listing, reading, creating, updating, and removing images from moodboards.
• Listing personalized profiles, folders, and storage information.
• Reading the public explore feed.
• Calling endpoints without a dedicated named tool through api_get and submit_raw_job.

The CLI adds login, setup diagnostics, browser-call capture, command discovery, help output, and output controls such as --json, --compact, and --select.

Limitations and notes

Midjourney MCP Server & CLI depends on a local signed-in browser, so a cloud-only connector, claude.ai, or a phone cannot reach the session on the developer’s machine. It is intended for local MCP clients such as Claude Code, Claude Desktop, and Cursor, as well as terminal workflows.

Image generation requires approval in the client or the CLI’s confirmation flow. The project does not provide a Midjourney API key abstraction; it operates through the browser session instead. Keep the managed browser profile available wherever the MCP server or CLI will run.