Skip to content

Summary

MiniMax's official MCP server for its media APIs: text to speech, voice cloning and voice design, image generation and editing, and Hailuo video generation.

Features

  • Text to speech with a voice catalogue, voice cloning and prompt-based voice design
  • Image generation with MiniMax-H3-Image at up to 4K, plus prompt-driven image editing
  • Hailuo-02 video generation at 768P or 1080P with async job polling
  • Outputs returned as URLs or saved to a local directory
  • Global and mainland-China API hosts; stdio and SSE transports

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it, filling in the environment variables with your own values.
  3. Restart the agent — the "MiniMax MCP" tools will be available.
{
  "mcpServers": {
    "minimax-mcp": {
      "command": "uvx",
      "args": [
        "minimax-mcp",
        "-y"
      ],
      "env": {
        "MINIMAX_API_KEY": "<YOUR_API_KEY>",
        "MINIMAX_MCP_BASE_PATH": "/path/to/output-dir",
        "MINIMAX_API_HOST": "https://api.minimax.io",
        "MINIMAX_API_RESOURCE_MODE": "url"
      }
    }
  }
}

Description

MiniMax MCP is the official server for MiniMax's generation APIs, letting an MCP client turn a request into audio, images or video without writing API code. Its tools are text_to_audio with list_voices, voice_clone from supplied audio files, voice_design (a new voice generated from a text description, with preview audio), text_to_image, image_edit, and generate_video with query_video_generation for polling long-running jobs.

Current models. As of the November 2026 release, text_to_image defaults to MiniMax-H3-Image with 1K/2K/4K output (the older image-01 stays available through the model parameter), and image_edit takes a prompt plus 1-16 reference images. Video generation supports MiniMax-Hailuo-02 with 6- or 10-second clips at 768P or 1080P. Because video jobs are slow, the README suggests telling the agent to submit the task and poll it rather than wait.

One configuration detail matters. MiniMax runs separate global and mainland-China platforms, and the API key must match the host — https://api.minimax.io or https://api.minimaxi.com — or every call fails with "invalid api key". Generated files are either returned as URLs or downloaded into MINIMAX_MCP_BASE_PATH, set by MINIMAX_API_RESOURCE_MODE. Calls are billed to your MiniMax account.

The server is Python, run with uvx minimax-mcp, over stdio or SSE; a JavaScript implementation (MiniMax-MCP-JS) is maintained separately. MiniMax now also points users to its mmx-cli command-line tool. MIT-licensed.

Featured

Replicate's official MCP server: search thousands of hosted models, read their schemas, and run predictions on image, video, audio and language models from inside an agent.

MCP: VoiceStudio

by Palash Deb

Local speech MCP server bundled with VoiceStudio — generate speech, clone a voice from reference audio, and transcribe in 646 languages, all on your own machine with no API key.

MCP: Deepgram

by Deepgram

Deepgram's official MCP server, bringing speech-to-text, text-to-speech and audio intelligence into Claude Code, Cursor and Windsurf.

Featured

Official MCP server for the Mux video API, built on a code-execution scheme: the agent writes TypeScript against the SDK and runs it in a Deno sandbox.

Browse all MCP servers →