Demo Ts MCP Client Server

🚀 MCP Basic Demo

This project demonstrates the basic client-server interaction using the Model Context Protocol (MCP). MCP allows AI models (such as those accessed via an AI SDK, like Google Gemini/Vertex AI) to securely discover and utilize external tools or resources provided by a separate server process.

In this example:

• The server (src/server/) acts as an MCP provider, offering simple computational tools (e.g., addition, subtraction).
• The client (src/client/) interacts with Google AI models using the AI SDK and connects to the MCP server, enabling the AI to use the server's tools during generation.

This setup shows how to extend the capabilities of an AI model by allowing it to access custom functions hosted on an MCP server.

🚀 Quick Start

Prerequisites

• Node.js (version >=23.0.0, as specified in package.json)
• npm (comes with Node.js)
• Google AI (Gemini) API Key. You can obtain it from Google AI Studio (required when using the Gemini model).

Installation

1. Clone the repository:

   git clone git@github.com:bertrandgressier/demo-ts-mcp-client-server.git
   cd demo-mcp-basic
   

2. Install dependencies:

   npm install
   

3. Create an environment file: Copy the example environment file .env.example to a new file .env:

   cp .env.example .env
   

   Then edit the .env file to add the actual API keys and configurations. The required and optional variables are as follows:

   # Required: for Google Studio models (Gemini)
   GOOGLE_API_KEY=YOUR_GOOGLE_API_KEY
   
   # Required: for Google Vertex AI models
   VERTEX_PROJECT_ID=YOUR_VERTEX_PROJECT_ID
   
   # Optional: defaults to 'us-central1' if not set
   # VERTEX_LOCATION=your-vertex-location
   
   # Optional: path to the Vertex AI service account key file. Defaults to 'vertex-key.json' in the root directory if not set.
   # VERTEX_KEY_FILE=path/to/your/vertex-key.json
   

   Replace YOUR_GOOGLE_API_KEY and YOUR_VERTEX_PROJECT_ID with your own values.

   ⚠️ Important Note

   Do not commit the .env file to the version control system.

Available Scripts

• npm run build: Compile the project source code to JavaScript.
• npm start: Run the compiled project.
• npm run dev: Run in development mode with automatic reload on changes.

Project Structure

• src/: Contains the TypeScript source code.
  • client/: Code for the MCP client application.
  • server/: Code for the MCP server application.
  • model.ts: Handles AI model initialization and environment variable loading.
• dist/: Compiled JavaScript code (generated after running npm run build).
• .env: Stores environment variables (API keys, etc.) - Do not commit this file.
• package.json: Project metadata and dependencies.
• tsconfig.json: TypeScript compiler configuration.

Documentation

This demo shows how to integrate external tools into an AI application using the MCP protocol. The server provides basic mathematical operation functions, and the client calls these services through interaction with Google Gemini.