ia-programming/youtube-mcp ? reverse-engineered prompt

Build me a Python MCP server for YouTube that I can plug into Claude Desktop or any MCP client.

I want it to let me search for YouTube videos, pull transcripts from a video URL, save the video info plus transcript into a vector database, and then do semantic search across the saved transcripts. It should work without using the official YouTube API for search, but it’s fine to use a Google API key for embeddings. Make it easy to run locally with uv, and include clear setup for a .env file and a simple way to start the server.

Please keep the tools straightforward and name them in a way that makes sense, so the client can discover them easily. If you need to check current docs or best practices for MCP or transcript access, look them up online. Also add basic error handling and helpful logging so it’s not annoying to debug when a video has no transcript or something fails.