A minimal Model Context Protocol server starter in TypeScript. Build custom AI tools in minutes.

MCP lets AI assistants like Claude and Cursor call your code — database queries, API calls, file operations, anything you can write a function for. This starter gives you a working server with everything wired up so you can skip the boilerplate and start building.

git clone https://github.com/tysoncung/mcp-server-starter.git
cd mcp-server-starter
npm install

Run in development mode (hot reload):

npm run dev

That's it. Your MCP server is running.

A single example tool (echo) that demonstrates the pattern for defining MCP tools:

server.tool(
  'echo',                                           // Tool name
  'Echo back the provided message',                 // Description for the AI
  { message: z.string().describe('Message text') }, // Input schema (Zod)
  async ({ message }) => {                          // Handler
    return { content: [{ type: 'text', text: `Echo: ${message}` }] };
  },
);

To add your own tools, follow the same pattern in src/server.ts. Define a name, describe what it does, declare the inputs with Zod, and write the handler.

Build the project first:

npm run build

Edit your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "starter": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-server-starter/dist/index.js"]
    }
  }
}

Restart Claude Desktop and your tools will be available.

Open Cursor Settings → MCP → Add Server, or edit .cursor/mcp.json in your project:

{
  "mcpServers": {
    "starter": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-server-starter/dist/index.js"]
    }
  }
}

src/
├── server.ts              # Server setup & tool definitions
├── index.ts               # Entry point (stdio transport)
└── __tests__/
    └── server.test.ts     # Tool tests

• Node.js ≥ 20
• npm

This free starter gets you up and running with one example tool and stdio transport. The Pro Starter Kit gives you everything you need to ship production MCP servers:

• 5 tool templates you can copy and customize — not just echo, but real patterns for databases, REST APIs, file systems, web scraping, and code execution
• Dual transport — serve over stdio for local use and SSE for remote clients, both from the same codebase
• Auth middleware — API key and JWT authentication out of the box
• Rate limiting — protect your server from runaway AI requests
• Structured logging — know exactly what your server is doing in production
• Docker + AWS CDK — deploy anywhere with one command
• Full test suite — every tool tested, CI-ready

Skip the hours of setup. Start building production MCP tools today.

MIT — see LICENSE

I Built a Production MCP Server Kit — Here's What I Learned — the design decisions, architecture, and lessons from building production MCP servers.