MCP-Template 🚀

A containerized Model Context Protocol (MCP) bridge for local and external LLMs. This project provides a robust, extensible hub that allows you to use your local models (via Ollama) or cloud models while giving them access to your local filesystem and web search tools.

web_search mocked fetch_private_data using api.example.com

🌟 Key Features

• Local Inference: Integrated Ollama service for running models locally (no internet required).

• Bridge Service: Python-based MCP server using FastMCP that acts as the tool hub.

• Filesystem Access: Secure access to your project files via mounted Docker volumes.

• Web Search: Extensible web search tool for real-time information.

• SSE Transport: Exposes the MCP server over HTTP (SSE) for easy integration with other apps.

Related MCP server: MCP Server

🛠️ Stack

• Orchestration: Docker Compose

• Inference: Ollama

• MCP Framework: FastMCP (Python)

• Networking: LAN-ready SSE (Server-Sent Events)

🚀 Getting Started

1. Configure Environment

Copy the template and fill in your API keys (optional for local-only use):

cp .env.template .env

2. Launch the Hub

docker-compose up -d --build

3. Verify Services

• Ollama: Accessible at http://localhost:11434

• MCP Bridge: Accessible at http://localhost:8000/sse

📂 Mounting Your Project

To give the LLM access to your specific project files, place them in the my_project_files/ directory, or modify the volumes section in docker-compose.yml:

volumes:
  - ./your_local_path:/data

🔗 Connecting to Claude Desktop

To use this as an MCP server in Claude Desktop, add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "omnimcp": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/inspector", "http://localhost:8000/sse"]
    }
  }
}

Note: This utilizes the MCP inspector to bridge the SSE connection to Claude Desktop's stdio-based expectations, or you can use other clients that support SSE directly.

🛠️ Customizing Tools

Add new tools in app.py by using the @mcp.tool() decorator.

📄 Licencia

Este proyecto está bajo la Licencia MIT - mira el archivo LICENSE para detalles.