TypeScript MCP Server Starter

A Model Context Protocol (MCP) server starter built with TypeScript, supporting Streamable-HTTP transport. This project uses Koa for the HTTP server implementation and provides a solid foundation for building MCP servers.

Inspired by: This starter project is inspired by the official MCP TypeScript SDK example from the Model Context Protocol team.

Features

• Streamable-HTTP Transport: Full support for MCP Streamable-HTTP protocol
• Koa-based HTTP Server: Modern, lightweight HTTP server using Koa and @koa/router
• Session Management: Full session lifecycle management with resumability support
• TypeScript: Fully typed with comprehensive type definitions
• Development Tools: Hot reloading, TypeScript compilation, and debugging support
• @modelcontextprotocol/sdk (v1.10.1) - Official MCP SDK for building protocol-compliant servers

Installation

This project uses Yarn as the package manager. To install dependencies:

yarn install

Building the Project

To build the TypeScript code:

yarn build

Running the Server

Production Mode

To run the server in production mode:

yarn start

The server will start on port 3000 by default. You can customize the port:

PORT=3002 yarn start

Development Mode

For development with hot reloading:

yarn dev

With a custom port:

PORT=3002 yarn dev

Testing with MCP Clients

1. Start the server

In one terminal, start your MCP server:

yarn start

2. Test with MCP Inspector

In a separate terminal, test your Streamable-HTTP MCP server using the official MCP Inspector:

npx @modelcontextprotocol/inspector http://localhost:3000/mcp

Or if you're using a custom port:

npx @modelcontextprotocol/inspector http://localhost:3002/mcp

The server runs on http://localhost:3000/mcp and supports the Streamable-HTTP transport protocol.

Optional: Deploy to Google Cloud Run

Want to deploy your MCP server to the cloud? This project includes everything needed for Google Cloud Run deployment:

Quick Deploy (3 steps):

1. Install and setup Google Cloud SDK:

   curl https://sdk.cloud.google.com | bash
   exec -l $SHELL  # or restart terminal
   gcloud auth login
   gcloud config set project YOUR_PROJECT_ID

2. Enable required APIs:

   gcloud services enable cloudbuild.googleapis.com run.googleapis.com containerregistry.googleapis.com

3. Deploy from Local:

   ./deploy.sh

That's it! The script will build, push, and deploy your server, then give you the public URL.

Alternative: Manual Deployment

# Build and push
docker build -t gcr.io/YOUR_PROJECT_ID/streamable-mcp-server .
docker push gcr.io/YOUR_PROJECT_ID/streamable-mcp-server

# Deploy
gcloud run deploy streamable-mcp-server \
  --image gcr.io/YOUR_PROJECT_ID/streamable-mcp-server \
  --platform managed --region us-central1 \
  --allow-unauthenticated --port 3000

For detailed deployment instructions, see the Deployment to Google Cloud Run section below.

Available Tools

The server comes with several example tools:

greet

A simple greeting tool that takes a name parameter.

Parameters:

• name (string): Name to greet

Example:

{
  "name": "greet",
  "arguments": {
    "name": "World"
  }
}

get_session

Gets the current session information.

Parameters: None

multi-greet

A demonstration tool that sends multiple notifications with delays.

Parameters:

• name (string): Name to greet

This tool demonstrates:

• Asynchronous operations
• Sending notifications during tool execution
• Multiple response phases

Environment Variables

Variable	Default	Description
PORT	3000	HTTP server port

Project Structure

├── src/
│   ├── index.ts          # Main entry point
│   └── server_runner.ts  # Server implementation with Koa
├── build/                # Compiled JavaScript output
├── package.json          # Project dependencies and scripts
└── tsconfig.json         # TypeScript configuration

Development

Adding New Tools

To add a new tool, modify the server setup in src/index.ts:

// Create the server
const { server, start } = createMcpServer({
  name: "your-server-name",
});

// Configure tools
server.tool(
  'my-tool',
  'Description of my tool',
  {
    param1: z.string().describe('Description of param1'),
    param2: z.number().describe('Description of param2'),
  },
  async ({ param1, param2 }, { sendNotification }): Promise<CallToolResult> => {
    // Tool implementation
    return {
      content: [
        {
          type: 'text',
          text: `Result: ${param1} and ${param2}`,
        },
      ],
    };
  }
);

// Start the server
const servers = start();

Troubleshooting

Common Issues

1. Port already in use: Change the PORT environment variable
2. TypeScript errors: Run yarn build to check for compilation errors
3. Client connection issues: Verify your MCP client is configured to use http://localhost:3000/mcp

Debugging

Enable verbose logging by setting the DEBUG environment variable:

DEBUG=* yarn start

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request