Creating Your First MCP Server

Setup the Environment

We'll build a simple TypeScript MCP Server that provides a "Weather" Tool to our LLM.

Step 1: Initialize Project

Create a new directory and initialize npm:

bash
mkdir mcp-weather-server
cd mcp-weather-server
npm init -y
npm install @modelcontextprotocol/sdk zod
npm install -D typescript @types/node tsx
npx tsc --init

Step 2: Write the Server Code

Create an index.ts file in your root folder:

typescript
import { Server } from "@modelcontextprotocol/sdk/server/index.js"
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js"
import { CallToolRequestSchema, ListToolsRequestSchema } from "@modelcontextprotocol/sdk/types.js"
import { z } from "zod"

// 1. Initialize Server Metadata
const server = new Server(
  {
    name: "weather-mcp",
    version: "1.0.0",
  },
  {
    capabilities: {
      tools: {},
    },
  },
)

// 2. Define the Tool (Instructing the LLM what it does)
server.setRequestHandler(ListToolsRequestSchema, async () => {
  return {
    tools: [
      {
        name: "get_weather",
        description: "Fetch the current weather for a specific city.",
        inputSchema: {
          type: "object",
          properties: {
            city: { type: "string", description: "City name e.g. Tokyo" },
          },
          required: ["city"],
        },
      },
    ],
  }
})

// 3. Handle the Execution (When the LLM calls the Tool)
server.setRequestHandler(CallToolRequestSchema, async (request) => {
  if (request.params.name === "get_weather") {
    const { city } = request.params.arguments as { city: string }

    // Hardcoded logic for simplicity
    const temp = Math.floor(Math.random() * 35)

    return {
      content: [{ type: "text", text: `The weather in ${city} is ${temp}°C and sunny.` }],
    }
  }

  throw new Error("Tool not found")
})

// 4. Start the Server Pipeline using Stdio
async function run() {
  const transport = new StdioServerTransport()
  await server.connect(transport)
  console.log("Weather MCP Server running on stdio")
}

run().catch(console.error)

Why Stdio?

The StdioServerTransport allows the MCP Client (like Claude Desktop) to start your MCP server seamlessly as a background process simply by invoking your script path. All communication happens invisibly over the standard input/output streams via JSON-RPC.

Test the Server Externally

Your server is fully built! You can run it via npx tsx index.ts. By itself, it won't do much because it expects external JSON-RPC requests on its standard input.

Testing with MCP Inspector

Anthropic provides a great tool to test your server without needing an AI. Run this in your terminal:

npx @modelcontextprotocol/inspector npx tsx index.ts

Next Steps

Continue Learning

Connect our newly created Weather server to Claude and LangChain so our AI can read the weather live!

Next Lesson

Integrating MCP with AI Agents