MCP Simply Explained

A gentle introduction to Model Context Protocol

zen

In this article we’ll use one screen to learn about the Model Context Protocol or MCP.

1. In the middle panel we have some code that we’ve added descriptive comments about what the code does and the parameters used to call the code.

2. We use code from a framework to turn our function into a server that follows the Model Context Protocol

   1. In this case we are using the Gradio framework, but other choices like Semantic Kernel or LLM specific libraries

3. In the bottom panel we are running our server.

4. In the left panel, the AI chat client or agent is told about available MCP servers

   1. In this case we are using Visual Studio Code, GitHub CoPilot and Claude 3.7 Sonnet

5. In the right pane,l the AI chat model is looking at available MCP servers, determining parameters, making a call and presenting results

In summary, MCP involves

1. Add descriptive comments to the code

2. Use a framework to create the MCP server

3. Run the server

4. Add MCP server details to chat clients

5. The AI chat model determines when and how to call a MCP server