Example MCP Server + Client Implementation

I wrote an article and created a video about MCPs and why they're useful in my opinion.

This demo project contains a backend service / API that's consumed by a MCP server which exposes it in a standardized way to MCP clients - like the example MCP client (a very simple AI chatbot) that's also part of this project.

Important: All three parts (service, server, client) rely on Node.js being able to execute TypeScript without a compilation step. Yes, modern Node.js can do that! Make sure you have Node.js 23.x or higher installed!

Configuration

Add a .env file inside the client folder (next to the package.json file there) and add the following content to it:

OPENAI_API_KEY=<your-open-ai-key>

In each folder (service, server, client) run npm install to install required dependencies.

Running Service, MCP Server & MCP Client

For each part, navigate into the respective folder (service, server and client) and run npm run dev (with Node.js 23+). Keep each process running.

The "client" process is a simply AI chatbot using OpenAI behind the scenes. This chatbot has the custom MCP server "installed" - therefore, this AI chat application is able to store and retrieve custom information via the backend service created in the "service" folder.

Bugs are possible - it's just a basic demo implementation :-)