
MiniMax MCP Server: Setup for Claude, Cursor and Other Apps
Lets Claude or Cursor make speech, clone voices, and generate images and videos with MiniMax models, using your MiniMax account.
This is MiniMax's own server. It runs on your computer with uvx and calls the MiniMax API, so you need an API key from your MiniMax account and some credit.
The key and the API address must match your region, global or mainland China, or you get an invalid key error. Most tools can cost money, and the tool text tells the assistant to use them only when you ask.
Good for
- Creators who want voice-overs and narration from a prompt
- Designers and marketers generating images or short videos
- Developers building voice features who want to test voices fast
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install MiniMax MCP
Claude Code
- 1
Copy this prompt.
Add the MiniMax MCP server to Claude Code. Run this command: claude mcp add minimax -e MINIMAX_API_KEY=<MINIMAX_API_KEY> -e MINIMAX_API_HOST=<MINIMAX_API_HOST> -- uvx minimax-mcp Ask me for MINIMAX_API_KEY, MINIMAX_API_HOST and only put the values in that config, nowhere else. Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires uv (Python).
Built from the package data we hold. Not yet tested by us.
Tools
Turns text into spoken audio
Converts text to speech with a MiniMax voice. By default it returns a link to the audio; with MINIMAX_API_RESOURCE_MODE set to local it saves the file to a folder (your Desktop by default). Speed, volume and other voice settings are optional. It can cost money.
Try: “Read this paragraph aloud in an audiobook voice and save it as an MP3.”
Write ·
text_to_audioGenerates images from a text prompt
Makes 1 to 9 images from a prompt with the image-01 model, with a choice of aspect ratio. By default it returns links; in local resource mode it saves the images to a folder. It can cost money.
Try: “Generate two 16:9 images of a lighthouse at dusk.”
Write ·
text_to_imageGenerates a short video from a prompt
Creates a video from a text or image prompt with MiniMax models, including camera-move instructions on the Director models. It can return a task ID to check later, and gives a video link, or saves the file only in local resource mode. It can cost money.
Try: “Make a 6 second video of waves hitting a rocky shore.”
Write ·
generate_videoChecks on a video that is still being made
Looks up the status of a video job by its task ID. When it is done you get a link, or in local resource mode the finished video is saved to a folder.
Try: “Is my video task finished? Save it when it is.”
Write ·
query_video_generationLists the voices you can use
Returns the available voices: system voices, your cloned voices, or both.
Try: “Which voices can I use for narration?”
Network ·
list_voicesClones a voice from an audio sample
Creates a new voice from an audio file on your computer or at a URL. It returns a demo clip as a link, or saves it in local resource mode. The README says the new voice is charged on first use.
Try: “Clone my voice from this recording and call it my-voice.”
Write ·
voice_cloneCreates a new voice from a text description
Generates a voice from a description such as 'a calm older narrator' and returns a preview of your sample text, as a link or a saved file in local resource mode. It can cost money.
Try: “Design a warm, friendly female voice and preview it saying 'Welcome back'.”
Write ·
voice_designPlays an audio file on your computer
Plays a WAV or MP3 file from a local path or a URL through your speakers. It does not play video.
Try: “Play the audio file you just generated.”
Files ·
play_audio
Read from the source code
Permissions
medium risk · 3 things to watch
Medium risk
Calls to the MiniMax API can use up your credit, and some tools write files to your computer. Give the assistant a key from an account with a small balance and a dedicated output folder.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
Tools can spend your MiniMax credit
Speech, image, video, voice cloning and voice design calls are billed by MiniMax, and a cloned voice is charged on first use. Use a key from an account with a small balance and watch your usage. github.com
Local resource mode writes files to your computer
By default the tools return links to the generated audio, images and video. If you set MINIMAX_API_RESOURCE_MODE to local, they download files to the output folder, which defaults to your Desktop. Set MINIMAX_MCP_BASE_PATH to a dedicated folder first. github.com
Voice cloning uploads your recording
voice_clone sends the audio file you point at to MiniMax. Only use recordings you have the right to clone. github.com
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Talks to | api.minimax.io or api.minimaxi.com (whichever host you set). URLs of audio files you give it |
|---|---|
| Sign-in and keys | The key sits in the MINIMAX_API_KEY environment variable in your client config. |
| Scopes |
|
Questions
Is the MiniMax MCP server official?
Yes. The README calls it the official MiniMax MCP server, and it lives in the minimax-ai GitHub organisation.
Do I need an API key?
Yes. Get one from the MiniMax platform, and set the matching API host: https://api.minimax.io for the global platform or https://api.minimaxi.com for mainland China. A mismatch gives an invalid API key error.
Is it free?
The server is free and MIT licensed, but its calls to MiniMax can cost money. The README warns that using the tools may incur costs.
What is the difference from the JavaScript version?
The README links MiniMax-MCP-JS as the official JavaScript implementation. This page covers the Python one, which you run with uvx.
Where do the generated files go?
Nowhere by default: the tools return links to the audio, image or video. Set MINIMAX_API_RESOURCE_MODE to local and the files are saved to the folder in MINIMAX_MCP_BASE_PATH, or your Desktop if that is unset.
Why does the server fail with spawn uvx ENOENT?
Your app cannot find uvx. Run which uvx in a terminal and put that full path in the command field of your config.
Can I use it in ChatGPT or n8n?
This page covers the local setup only. The README supports stdio, and also SSE for hosted deployments, but it gives no steps for ChatGPT or n8n.
Information
updated Nov 2, 2026
- Publisher
- Official
- Package
minimax-mcp- Registry
- PyPI
- Transport
- stdio
- Auth
- API key
- License
- MIT
- Last commit
- Aug 20, 2026
- Open issues
- 21
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.