MiniMax MCP

What it does

MiniMax's official MCP server exposes Hailuo video generation, text-to-image, and best-in-class multilingual TTS with voice cloning and voice design. One vendor and one key covers both your shots and your dialogue — rare breadth for a single official server. Music generation was part of that breadth until August 2026, when the tool was retired.

Connect

claude mcp add minimax -e MINIMAX_API_KEY=YOUR_KEY -e MINIMAX_API_HOST=https://api.minimax.io -- uvx minimax-mcp

{
  "mcpServers": {
    "minimax": {
      "command": "uvx",
      "args": [
        "minimax-mcp"
      ],
      "env": {
        "MINIMAX_API_KEY": "YOUR_KEY_HERE"
      }
    }
  }
}

{
  "mcpServers": {
    "minimax": {
      "command": "uvx",
      "args": [
        "minimax-mcp"
      ],
      "env": {
        "MINIMAX_API_KEY": "YOUR_KEY_HERE"
      }
    }
  }
}

Sample tools

Field notes

MIT, 1,590 stars, with an official JS sibling (MiniMax-AI/MiniMax-MCP-JS, 130 stars, pushed 2026-08-21). Re-checked 2026-10-17: no commits on either repo since August and PyPI is still 0.0.19 — quiet, not abandoned, and the vendor's MCP guide still documents both servers. The three-month quiet spell ended on 2026-08-20: the server migrated to MCP Python SDK v2 and **retired the `music_generation` tool** in the same batch, shipped as PyPI minimax-mcp 0.0.19 on 2026-08-21. Score elsewhere — Mureka, Sonilo or Epidemic Sound. `play_audio` survives in the code but has been dropped from the README's tool table, so treat it as undocumented rather than guaranteed. Watch the model drift: the default text-to-video model constant is now `MiniMax-Hailuo-2.3`, while `generate_video`'s own tool description still enumerates only up to `MiniMax-Hailuo-02` and the older T2V/I2V-01 Director line — an agent reading the schema will not know what it is actually being given. Default speech model is `speech-2.8-hd`. image_to_video is JS-only. The README now steers you to MiniMax's `mmx-cli` agent skill ahead of this server, which is the vendor telling you where the newest models land first. MINIMAX_API_RESOURCE_MODE toggles URL vs local-file output; MINIMAX_MCP_BASE_PATH sets the output directory.

Also in Video Generation

Magic Hour MCPMagic Hour Official

Magic Hour's hosted MCP: 44 tools across video, image and voice generation on one connection.

text-to-video · image-to-videoremote · credits

Google GenMedia MCPGoogle Cloud Official

Google Cloud's MCP servers for Veo, Gemini Omni, Lyria, Chirp TTS and ffmpeg compositing.

text-to-video · image-to-videopaid

Runway MCPRunway Official

Official hosted MCP: Gen-4.5, Seedance, Kling and Veo video from any agent.

text-to-video · image-to-videoremote · credits

C Clip Studio MCPClip Studio Official

Hosted MCP for a story-to-video studio: multi-shot narrated films, labs, and shorts.

text-to-video · image-to-videoremote · freemium

Ark Seed Multimodal MCPBytePlus SA program (byteplus-sa) — vendor-adjacent, not the ModelArk product team

Seedance video, Seedream stills, and Seed Audio behind one typed BytePlus MCP server.

text-to-video · image-to-videopaid

Luma API MCPLuma AI Official

Luma's own two-tool MCP server for Ray video and Photon stills.

text-to-video · image-to-videopaid

Pairs well with

AKOOL MCPAKOOL Official

Hosted MCP for AKOOL face swap, character swap, image-to-video, TTS and avatar lookup.

face-swap · character-swapremote · credits

Apatero Studio MCPApatero Official

Studio MCP for talking avatars, locked recurring characters, and batched output.

talking-head · avatar-videoremote · credits

Artlist MCPArtlist Official

Official hosted Artlist MCP — generate images, video, music and voice-over on your Artlist credits.

text-to-image · text-to-videoremote · credits