Practice MCP JS
A Model Context Protocol (MCP) server implementation in TypeScript that provides weather information tools for AI assistants.
Description
This project demonstrates how to create an MCP server that exposes weather-related tools to Large Language Models (LLMs). The server provides a fetch-weather tool that retrieves current weather data and forecasts for any city using the Open-Meteo API.
Features
- Weather Tool: Fetch current weather conditions and forecasts for any city
- Input Validation: Uses Zod for robust input/output validation
- MCP Protocol: Compatible with AI assistants that support the Model Context Protocol
- TypeScript: Fully typed implementation for better development experience
Prerequisites
- Node.js (version 14 or higher)
- npm or yarn package manager
Installation
- Clone the repository:
git clone <repository-url>
cd practice-mcp-js
- Install dependencies:
npm install
Usage
Running the Server
To start the MCP server:
npx tsx main.ts
The server will start and listen for MCP connections via stdio transport.
Available Tools
fetch-weather
Retrieves weather information for a specified city.
Parameters:
city(string): Name of the city to get weather information for
Example Response: The tool returns current weather conditions including:
- Temperature
- Precipitation
- Day/night status
- Rain information
- Hourly forecast data
Project Structure
practice-mcp-js/
├── main.ts # Main server implementation
├── package.json # Project configuration and dependencies
├── package-lock.json # Dependency lock file
└── README.md # This file
Dependencies
- @modelcontextprotocol/sdk: Core MCP SDK for server implementation
- zod: Schema validation library for input/output validation
API Integration
This project integrates with the following external APIs:
- Open-Meteo Geocoding API: For converting city names to coordinates
- Open-Meteo Weather API: For retrieving weather data and forecasts
How It Works
- The server creates an MCP server instance with the name "PRO INDUSTRIAL MCP SERVER"
- It defines a
fetch-weathertool that:- Accepts a city name as input
- Geocodes the city name to get coordinates
- Fetches weather data using the coordinates
- Returns formatted weather information
- The server connects via stdio transport for communication with AI assistants
Development
The main server logic is implemented in main.ts:7-55. Key components include:
- Server initialization (
main.ts:7-10) - Tool definition (
main.ts:13-50) - Transport setup (
main.ts:54-55)
Error Handling
The weather tool includes error handling for:
- Cities that cannot be found
- API request failures
- Invalid input parameters (via Zod validation)
License
ISC
Contributing
This is a practice project for learning MCP server development. Feel free to fork and experiment with additional tools and features.