Gujejiani/mcp-server

Weather API & MCP Server

This project is a Clean Architecture Node.js/TypeScript application that exposes weather information via both a REST API and the Model Context Protocol (MCP) for LLM agent integration.

Features

• REST API: /weather/current returns static weather data (Tbilisi, 22°C, Clear)
• MCP Server: Exposes a getCurrentWeather tool for LLM agents via stdio
• Clean Architecture: Domain, Application, Infrastructure, and API layers
• Structured Logging: Pino logger, contextual fields, no sensitive data
• Dependency Injection: Typedi
• TypeScript, ESLint, Prettier: Enforced code quality

Project Structure

<repo>
├── src/
│   ├── api/           # Express REST API, controllers, routes
│   ├── application/   # Use cases, interfaces
│   ├── domain/        # Models, errors
│   ├── infrastructure/# DB/API implementations
│   └── mcpServer.ts   # MCP server entry point
├── tests/             # Unit and integration tests
├── package.json
├── tsconfig.json
└── README.md

Running the REST API

npm run build
npm start
# Visit http://localhost:3000/weather/current

Running the MCP Server (for LLM agents)

MCP_MODE=true npm run dev

• The MCP server will start in stdio mode.
• Connect an LLM agent (e.g., Claude Desktop, MCP Inspector) with the launch command:

  MCP_MODE=true node dist/api/server.js
  

• The agent can call the getCurrentWeather tool with { city: "Tbilisi" }.

Tool Example

• Tool Name: getCurrentWeather
• Input: { city: string }
• Output:

  {
    "temperature_c": 22,
    "condition": "Clear",
    "city": "Tbilisi",
    "timestamp": "2025-05-12T12:00:00.000Z"
  }
  

Notes

• All logs are sent to stderr to avoid interfering with MCP protocol messages.
• Weather data is static for demonstration. Replace with real API integration as needed.
• See copilot-instructions.md for full coding and architecture guidelines.