Creating the MCP Server

Let’s begin our hands-on journey into building an MCP application. We’ll create a tool-based server that allows AI systems to interact with external knowledge sources. We aim to expose well-defined capabilities that an LLM-powered agent can invoke securely, modularly, and contextually. To keep things practical, we’ll walk through the implementation using a scenario that reflects how real users might interact with such a system.

Scenario: Wikipedia research assistant

Imagine we’re building a knowledge assistant for curious minds—students, journalists, or professionals—who frequently use Wikipedia for background research. Instead of opening tabs and scanning walls of text, users should be able to ask “Tell me about Alan Turing,” or “Summarize the topic of carbon cycles,” and receive structured results: a clean summary, the article title, and a link to the full page.

For this purpose, we will use the MCP to communicate with a backend server that understands how to retrieve and package information from Wikipedia. In this lesson, we’ll build that server component of such a system using Python. Let’s begin by preparing the development environment.

Setting up the environment

Before we implement our server, we need to set up the environment. The following are the required libraries:

• wikipedia: A simple Python wrapper for the Wikipedia API. We’ll use it to search topics, fetch article metadata, and extract content.

• mcp: The official Python SDK for Model Context Protocol. It provides the FastMCP interface that lets us expose tools from our server.

We can install these libraries locally using the following commands:

Note: You don’t need to worry about these installations in the course. We have already set up the environment for you. You can focus directly on writing and executing the code.

Implementing the fetch_wikipedia_info tool

Now we are ready to implement the tools that will perform actions on the server side. We will start ...