Alpic

You can run an MCP server template that lets you build and expose tools and prompts to AI assistants and other MCP clients. It provides a simple HTTP transport path and a local development workflow so you can test, iterate, and extend your MCP-based integrations quickly.

How to use

Install and run the server locally, then connect an MCP client to the main MCP endpoint at your development URL. You can expose your own tools and prompts by adding them to the server configuration, and you can test interactions with the built‑in inspector tool. Use the HTTP transport to communicate with AI assistants and other MCP clients and send messages through the /mcp endpoint.

In practice, you will start the server in development mode to enable hot-reload, make changes, and see updates immediately. You can also build a production-ready bundle for deployment and run an inspector to validate requests and responses. The server exposes a primary endpoint for MCP communication at POST /mcp and will reject other HTTP methods on that path with a 405 error, guiding you to use the correct entry point.

How to install

Prerequisites: Node.js 22+.

Step-by-step setup:

# 1) Install Node.js 22+ (use the version configured for the project)
# 2) Clone the template repository
git clone <repository-url>
cd mcp-server-template

# 3) Install dependencies
npm install

# 4) Create environment file
cp .env.example .env

Additional content

Development commands you will use regularly include starting the server with hot-reload, building for production, and inspecting MCP interactions.

Development workflow commands shown in the template are:

• Start development server with hot-reload
• Build for production
• Run the MCP inspector to test interactions

npm run dev
npm run build
npm run inspector

Available tools

tool-name

Tool description that demonstrates how to add a new tool to the MCP server by defining parameters and a simple execution function.

prompt-name

Prompt description showing how to declare a prompt template with parameters and a handler that returns a user-facing message.