mcp

A simple project demonstrating how to build and use MCP (Model Context Protocol) servers.

1. Install dependencies:

pip install -r requirements.txt

1. Configure MCP server in Windsurf/Claude Desktop:

{
  "mcpServers": {
    "github": {
      "command": "python",
      "args": ["/path/to/server/github.py"],
      "env": {
        "GITHUB_TOKEN": "your_github_token_here (see .env.example)"
      }
    },
    "database": {
      "command": "python", 
      "args": ["/path/to/server/database.py"]
    }
  }
}

1. Use with LLM:

• Start Windsurf/Claude Desktop
• Ask: "Show me octocat's repositories"
• LLM will automatically call the MCP server tools

This MCP server is deployed and available for testing at:

• URL: https://my-mcp.fastmcp.app/mcp
• Status: Production ready

Configure Windsurf to use the deployed endpoint:

{
  "mcpServers": {
    "github-mcp-deployed": {
      "url": "https://my-mcp.fastmcp.app/mcp",
      "disabled": false
    }
  }
}

Test the deployed server:

• Ask any LLM with MCP support: "Show me repositories for gn1264"
• The deployed server will respond with GitHub data

• ✅ No local Python environment required
• ✅ Always available and reliable
• ✅ Managed infrastructure
• ✅ Same functionality as local version

For development and testing:

# Test GitHub server
python testing/test_github_server.py get-repos octocat

# Test Database server
python testing/test_database_server.py query "SELECT * FROM users"

mcp/
├── server/                 # MCP servers
│   ├── github.py          # GitHub API server
│   └── database.py        # Database server
├── client/                 # Real LLM client
│   └── mcp_client.py      # LLM integration example
├── testing/               # Server testing utilities
│   ├── test_gith