AI Toolbox MCP Server (ai-toolbox-mcp)

Wraps the 22 AI interfaces of Wanhe AI Tools into a standard MCP (Model Context Protocol) Server, so AI agents like Claude Desktop / Cursor can directly call local AI capabilities.

✨ Features

• Local inference: Based on local Ollama (qwen3 series), zero API costs

• Data stays on your machine: All inference runs locally, privacy-safe

• 22 interfaces: Summarization/translation/review/classification/RAG/comparison/batch/sentiment/FAQ/diff/meeting minutes/data extraction/paraphrasing/chat, etc.

• Plug and play: Configure once and it can be discovered and called by Claude Desktop / Cursor

Related MCP server: apple-fm-mcp

📦 Installation

pip install mcp fastmcp httpx requests
python mcp_server.py

If the mcp library is not installed, the script automatically falls back to the built-in stdio JSON-RPC implementation (same functionality).

⚙️ Claude Desktop Configuration

Edit claude_desktop_config.json:

{
  "mcpServers": {
    "ai-toolbox": {
      "command": "python",
      "args": ["/path/to/mcp_server.py"]
    }
  }
}

After restarting Claude Desktop, you will see the 17 AI tools from ai-toolbox in the tool list.

🛠 Supported 22 Interfaces

AI tools (17): summarize / translate / review / classify / rag_query (knowledge base Q&A) / rag_upload (knowledge base upload) / compare / batch / keywords / batch_summarize (batch summarization) / sentiment (sentiment analysis) / faq_generate (FAQ generation) / diff / meeting_minutes (meeting minutes) / data_extract (data extraction) / paraphrase / chat

System interfaces (5): health / pricing / stats / orders / root

💰 Pricing

• MCP Server subscription: ¥49/month

• Subscribe at https://wanhetools.top/developer, or choose a service at https://wanhetools.top/pay

• Local inference with zero API call fees

📄 Files

• mcp_server.py — MCP Server main program

• mcp_config.json — Configuration file (toolbox address/port/installation guide)

• README.md — This documentation