MCP Server

A TypeScript implementation of a Model Context Protocol (MCP) server, designed to evaluate MCP capabilities over/with RAG (Retrieval-Augmented Generation) approaches for Nutrimate and other AI applications.

Overview

This project explores the Model Context Protocol as an alternative to RAG systems, providing a standardized way for AI applications to access external data sources and tools. The server is built with TypeScript, Node.js and the MCP TypeScript SDK.

Features

• TypeScript-first: Built with full TypeScript support for better developer experience
• MCP Protocol: Implements the Model Context Protocol via Anthropic's TypeScript SDK

Prerequisites

• Node.js (v18 or higher)
• npm or yarn
• TypeScript knowledge (recommended)

Getting Started

1. Clone the repository:

   git clone https://github.com/samrasugu/mcp-server.git
   cd mcp-server
   

2. Install dependencies:

   npm install
   

3. Development mode:

   npm run dev
   

4. Build the project:

   npm run build
   

5. Start the production server:

   npm run start
   

Environment Variables

Set the following environment variables (or configure them in your MCP file):

• GOOGLE_PUBLIC_API_KEY: Your Google API key
• GOOGLE_CALENDAR_ID: Your Google Calendar ID (e.g., your Gmail address)

You can set these in a .env file:

GOOGLE_PUBLIC_API_KEY=your-google-api-key
GOOGLE_CALENDAR_ID=your-calendar-id

Usage

Get Today's Meetings

To print today's meetings directly in your terminal:

node src/index.js today

Integrating with Cursor MCP & Chat Interaction

This project supports integration with Cursor MCP, enabling you to interact with your server via chat and automate workflows.

1. Configure MCP Server in Cursor

1. Open Cursor and go to the MCP panel.
2. Add a new MCP server with the following settings (example):
   • Name: [your name]'s Calendar
   • Command: node
   • Args: src/index.ts
   • Host: localhost
   • Port: 3000
   • Environment Variables: Set GOOGLE_PUBLIC_API_KEY and GOOGLE_CALENDAR_ID as needed.

You can also create a .cursor/mcp.json file for quick setup.

The content of the .cursor/mcp.json file should be like this:

{
  "servers": [
    {
      "name": "[your name]'s Calendar",
      "command": "node",
      "args": ["src/index.ts"],
      "host": "localhost",
      "port": 3000,
      "env": {
        "GOOGLE_PUBLIC_API_KEY": "your-google-api-key",
        "GOOGLE_CALENDAR_ID": "your-calendar-id"
      }
    }
  ]
}

2. Start the MCP Server

Run the server locally:

node src/index.ts

Or use the MCP panel's built-in controls to start/stop the server.

3. Interact via Chat

• Open the chat panel in Cursor.
• Select your MCP server (e.g., "Sam's Calendar") from the chat source dropdown.
• Ask questions like:
  • Do I have any meetings today?
  • What is my next event?
  • List all meetings for this week.
• The server will respond with information from your Google Calendar.

4. Customization

You can extend the server to support more commands or integrate with other tools. Update src/index.ts and restart the server to apply changes.

Project Structure

mcp-server/
├── src/
│   └── index.ts          # Main server implementation
├── package.json          # Dependencies and scripts
├── tsconfig.json         # TypeScript configuration
└── README.md            # This file

Development

This project is actively being developed as part of an evaluation comparing MCP to RAG systems. The implementation may evolve as the evaluation progresses.

Available Scripts

• npm run dev - Start development server with hot reload
• npm run build - Build the TypeScript project
• npm run start - Start the production server
• npm test - Run tests (to be implemented)

MCP vs RAG Evaluation

This server is being developed to evaluate the effectiveness of the Model Context Protocol compared to traditional RAG approaches for AI applications, particularly focusing on:

• Performance: Response times and resource usage
• Flexibility: Ease of adding new data sources and tools
• Maintainability: Code organization and extensibility
• Integration: How well it works with existing AI workflows

License

MIT