CbetaMCP

# 🧠 MCP Aggregated Tool Service     A modular, extensible and FastAPI-based MCP (Multi-Component Platform) tool aggregation service. Easily connect and expose independent tools through standardized APIs. Perfect for frontend integration or large language model (LLM) orchestration. --- ## 🌟 Features - ✅ Automatic registration of multiple tools (based on directory scanning) - ✅ All interfaces use unified POST method + BaseModel validation - ✅ Asynchronous httpx interface calls supported - ✅ Standardized JSON response format (success/error) - ✅ One-click Docker deployment support - ✅ Accompanying development documentation for easy expansion of tool modules --- ## 📁 Project Structure ```bash . ├── main.py # FastAPI main program, including MCP registration logic ├── tools/ # Tool directory, one function per file ├── Dockerfile # For building images ├── docker-compose.yml # One-click deployment support ├── mcp_tool_开发说明.md # Developer usage specification document (Chinese) └── README.md ``` --- ## 🚀 Quick Start ### 🧰 Requirements - Python 3.8+ - pip - Docker / Docker Compose (optional) ### 📦 Local Dev ```bash # Install dependencies pip install -r requirements.txt # Start service python main.py ``` Default service address: http://localhost:8000/mcp ### 🐳 Docker Deployment ```bash # Build & Run docker-compose up --build -d # Access MCP tool service http://localhost:8000/mcp ``` --- ## 🧱 Tool Module Guidelines Each tool should be placed in the `tools/` directory (can be nested multiple layers) and include: 1. Define parameters using `pydantic.BaseModel`; 2. Register the tool function using `@__mcp_server__.tool()`; 3. Return `success_response()` or `error_response()`; 4. (Optional) Asynchronously call external interfaces + cache results. Example reference: ```python from pydantic import BaseModel from main import __mcp_server__, success_response class MyParams(BaseModel): name: str @__mcp_server__.tool() async def hello_tool(params: MyParams): return success_response({"message": f"Hello {params.name}!"}) ``` --- ## 🔗 API Usage All tool interfaces are accessed through the `/mcp` path, automatically registered according to the module. Request method: `POST` Request format: `application/json` Response format: ```json { "status": "success", "result": { ... } } ``` --- ## 📚 Docs - [📘 开发说明文档 (中文)](./mcp_tool_开发说明.md) - Example module path: `tools/xxx/tool_name.py` --- ## 📄 License MIT License © 2025 [your-name]