MiniMax Unified MCP

MiniMax Unified MCP

Unifies MiniMax's multimodal generation, web search, image understanding, audio, video, and music tools into a single MCP server for use with Claude and other clients.

Category
访问服务器

README

<div align="center">

MiniMax Unified MCP

⚡ Unified MiniMax MCP for multimodal generation, Token Plan search, and image understanding.

MIT License Python Tests MCP MiniMax

English | 简体中文

</div>

✨ Overview

minimax-unified-mcp merges three capability groups into one installable MCP server:

  • Official multimodal MiniMax generation tools
  • Official Token Plan tools: web_search and understand_image
  • Token Plan quota lookup aligned with the ExoMind Team workflow

This repository is intended for local-first MCP usage in Claude, Codex, and compatible MCP clients.

🧰 Tools

Tool Purpose
quota_tool Query Token Plan quota and refresh window
web_search_tool Search the live web via Token Plan
understand_image_tool Analyze a local image path, remote URL, or data: URL
text_to_audio_tool Generate speech with configurable output mode
text_to_audio_streaming_tool Low-latency speech playback path
list_voices_tool List available voices
voice_clone_tool Clone a voice from sample audio
play_audio_tool Play local or remote audio
generate_video_tool Generate text-to-video or image-to-video
query_video_generation_tool Query video generation status
text_to_image_tool Generate images
music_generation_tool Generate music from prompt and lyrics
voice_design_tool Design a new voice

🚀 Quick Start

1. Install

cd path/to/minimax-unified-mcp
python -m pip install -e ".[dev]"

2. Configure environment

Copy .env.example to .env and fill in your credentials.

Important variables:

  • MINIMAX_TOKEN_PLAN_API_KEY: Token Plan API key
  • MINIMAX_API_HOST: https://api.minimax.io or https://api.minimaxi.com
  • MINIMAX_MCP_BASE_PATH: base path for local artifacts
  • MINIMAX_API_RESOURCE_MODE: url or local
  • FASTMCP_LOG_LEVEL: logging level, usually WARNING or INFO

3. Add to your MCP client

{
  "mcpServers": {
    "MiniMaxUnified": {
      "command": "python",
      "args": ["-m", "exomind_minimax_mcp"],
      "env": {
        "MINIMAX_TOKEN_PLAN_API_KEY": "YOUR_TOKEN_PLAN_KEY",
        "MINIMAX_API_HOST": "https://api.minimax.io",
        "MINIMAX_MCP_BASE_PATH": "./output/minimax",
        "MINIMAX_API_RESOURCE_MODE": "local",
        "FASTMCP_LOG_LEVEL": "WARNING"
      }
    }
  }
}

See mcp_server_config_demo.json for a ready-to-copy example.

🧭 Recommended Usage

Web search

Use web_search_tool for current external information.

{
  "query": "MiniMax Token Plan latest image model"
}

Image understanding

Follow the official Token Plan MCP signature: prompt + image_source.

image_source supports:

  • local file path, such as D:/images/demo.png
  • HTTP / HTTPS image URL
  • data: URL
{
  "prompt": "Describe the UI structure in this screenshot",
  "image_source": "D:/images/screenshot.png"
}

Image generation then image understanding

  1. Call text_to_image_tool
  2. Copy one returned image URL
  3. Pass that URL into understand_image_tool.image_source

Low-latency TTS

Prefer text_to_audio_streaming_tool when perceived latency matters.

{
  "text": "hello from MiniMax"
}

Use text_to_audio_tool when you need more control over resource_mode, local output, or autoplay.

Quota field notes

quota_tool currently maps the upstream fields to their display semantics as follows:

  • current_interval_usage_count: remaining in current interval(当前窗口剩余)
  • current_weekly_usage_count: weekly remaining(本周剩余)

Note: even though the field name contains usage_count, this output is treated as remaining quota, not used quota.

Video generation

  • Default text-to-video path uses MiniMax-Hailuo-2.3
  • If you choose MiniMax-Hailuo-2.3-Fast, also provide first_frame_image
  • first_frame_image supports local path, URL, or data: URL
{
  "prompt": "A cute orange cat sleeping in sunlight",
  "model": "MiniMax-Hailuo-2.3-Fast",
  "first_frame_image": "D:/images/cat.png",
  "async_mode": true
}

Music generation

Music generation is slower than typical text APIs. The unified client uses an extended request timeout for this endpoint, but real latency still depends on upstream service status.

⚙️ Configuration

Detailed guides:

📦 Output Modes

Two output modes are supported:

  • url: return remote URLs directly
  • local: download artifacts to MINIMAX_MCP_BASE_PATH

Use local when you want reproducible local files for downstream automation.

🛡️ Errors and Quota Signals

The unified client preserves upstream errors and improves readability for common account states:

  • 1008 insufficient balance: clearly reported as balance insufficient
  • 2056 usage limit exceeded: clearly reported as quota / usage exhausted
  • 2013 invalid params: preserved for bad payloads such as unsupported video mode combinations

✅ Testing

Run the full suite:

python -m pytest -v

Run the live API matrix:

python scripts/run_live_api_matrix.py --json

The live matrix classifies results into statuses such as passed, unsupported, insufficient_balance, usage_limit_exceeded, timeout, and invalid_params.

🌍 Open Source

This repository is released under the MIT License.

📌 Status

The package runs from src/exomind_minimax_mcp/ and no longer depends on uvx for local development. It is ready for GitHub-based release packaging and ongoing MCP client integration.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选