Practical MCP with FastMCP & LangChain

Testing the Workflow

The questions below are designed to walk through each feature of the stack in order of increasing complexity. Type them at the Netflix> prompt one at a time.

The first two questions exercise the get_top_movies tool and the progress handler. You should see [Progress] lines printed as each result is processed, followed by a ranked list in the final answer.

What are the top 10 most viewed movies?

Show me the top 3 movies by hours viewed

This question tests search_movies. Every match triggers the elicitation handler: a single result asks you to confirm it is the right movie (yes/no), while multiple results present a numbered list and ask you to pick one.

What are the global viewing stats for 'Leave the World Behind'?

The next two questions test session state. The first call writes a new entry to the Redis-backed favorites list; the second reads it back. Both should reflect the same data because the entire REPL session runs inside a single open mcp_client connection.

Add 'Leave the World Behind' to my favorites

Show me my favorites list

This question chains two tool calls in one turn — the LLM first calls get_top_movies