MCP Weather Server

MCP Weather Server

A TypeScript server implementation for the Model Context Protocol (MCP) using Server-Sent Events (SSE) transport to provide weather information tools.

Overview

This server implements the Model Context Protocol and exposes weather forecast and alert tools through Server-Sent Events (SSE) transport. Unlike standard examples that typically use stdio for local communication, this server demonstrates how to implement remote connections via HTTP, allowing clients to connect over a network.

Features

• Implements MCP over SSE transport
• Provides two weather tools:
  • get-forecast: Get weather forecasts for specific coordinates
  • get-alerts: Get active weather alerts for US states
• Connects to the US National Weather Service API
• Supports multiple client connections

Prerequisites

• Node.js (v16 or newer)
• npm or yarn

Installation

# Clone the repository
git clone https://github.com/tyfeng1997/mcp-server.git
cd mcp-server

# Install dependencies
npm install

# Build the TypeScript code
npm run build

Usage

To start the server:

node build/index.js

This will start the server on port 3001 by default. You should see output similar to:

Weather MCP Server running on http://localhost:3001
Connect clients to http://localhost:3001/sse

Available Tools

The server exposes two tools:

1. get-forecast

   • Description: Get weather forecast for a location
   • Parameters:
     • latitude (number): Latitude of the location (-90 to 90)
     • longitude (number): Longitude of the location (-180 to 180)
   • Note: Only US locations are supported by the NWS API

2. get-alerts

   • Description: Get weather alerts for a state
   • Parameters:
     • state (string): Two-letter US state code (e.g., CA, NY)

Testing the Server

You can test if the server is running correctly by making a request to the SSE endpoint:

curl http://localhost:3001/sse

This should return an event with a session ID.

Companion Client

This server is designed to work with the MCP client available at tyfeng1997/mcp-client. The client connects to this server and allows AI assistants like Claude to use the weather tools through natural language.

Customization

Port Configuration

To change the port, modify the PORT constant in the code:

const PORT = 3001; // Change to your desired port

Adding More Tools

You can add more tools by following the pattern used for the existing tools:

server.tool(
  "your-tool-name",
  "Your tool description",
  {
    // Parameters schema using zod
    param1: z.string().describe("Parameter description"),
    param2: z.number().describe("Parameter description"),
  },
  async ({ param1, param2 }) => {
    // Tool implementation
    return {
      content: [
        {
          type: "text",
          text: "Tool result",
        },
      ],
    };
  }
);

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.