MCP Fundamentals

In this video, you will see how MCP Prompts function as a bridge between user input and the LLM. You will use the Inspector to fill out the travel itinerary template you just created.

First, make sure to stop any previous servers. Then, run your prompts server script:

uv run --with mcp prompts_demo_server.py

With the server running on port 8000, launch the Inspector (or refresh your existing session) and ensure it is connected to http://localhost:8000/mcp.

Navigate to the Prompts tab.

Click List Prompts. You should see the draft_itinerary prompt available. Click on it to expand the details.

Unlike tools (which run code) or resources (which fetch data), this interface represents a form for the user. You will see fields for destination and days.

Enter “London” for the destination and “3” for the days.

Click Get Prompt.

The Inspector sends these inputs to your Python server. The server injects them into the string template and returns the result.

Look at the output on the right. You will see a JSON object containing a messages array. Inside, there is a role “user” with the content: “Please create a detailed 3-day itinerary for a trip to London…”

This demonstrates exactly what the Host Application does: it takes simple form inputs from the user, generates a complex, well-structured prompt using your server, and prepares it to be sent to the LLM context window.