MCP Elicitation Tutorial
Learn how to build interactive MCP tools.
MCP elicitation is new in the latest version of the spec. It allows for user-input workflows like:
Confirmation (y/n)
Multiple choice selection
Freeform input
I’ll show you how it works and how to build MCP servers with elicitation today using FastMCP.
Check out the video on YouTube
Timestamps
1:10 - How MCP Elicitation Works
4:20 - MCP Elicitation Demo
10:21 - MCP Elicitation with FastMCP
16:16 - Elicitation with the Model Context Protocol Inspector
21:12 - Elicitation with VS Code
Massive Course Update 🚀
I’ve added 32 videos to the AI Engineer Roadmap, over 10 hours of content.
Have a look at the first lesson video for free and consider buying the full course if you like it!
Topics from this week’s video
MCP Client Feature: Elicitation
Demonstration of a new MCP client capability enabling user input mid-process
Tested using Claude Desktop, VS Code Insiders, and MCP Inspector
How Elicitation Works
Three-step flow: server request → user prompt → response returned
Allows structured data input (e.g., name, email, choices)
Implementation Example
Python server built with FastMCP (“Book Buy Demo”)
Tools:
list_booksandbuy_bookdemonstrating elicitation stepsShows async
context.elicit()usage and response handling
Testing & Debugging
Claude Desktop: no elicitation support yet
VS Code Insiders & MCP Inspector: successful interactive flow
Fixes for response handling and validation
Practical Use & Takeaways
Use elicitation to confirm actions or gather info before running tools
Found via GPT-generated FastMCP code
Encouragement to explore MCP features and related Zen Codes videos
First-principle thinking is required to break out of iterative improvements and make order of magnitude improvements.
Simon Eskildsen


