YouTube Transcript MCP

This MCP server does one thing: it pulls transcripts from YouTube videos. You give it a URL, it returns timestamped text. It handles multiple URL formats (watch, youtu.be, embed) and supports language selection when multiple captions are available. No API key needed, which is the main reason I'd reach for it over the official YouTube Data API. I've used it for summarising conference talks and extracting quotes from interviews. The workflow is simple: paste a YouTube link into Claude, ask for a summary or specific sections, and the MCP fetches the transcript in the background. It's faster than scrubbing through video manually, and the timestamps let you jump back to context when needed. For content repurposing, it's solid. I've turned hour-long talks into blog outlines in minutes. The tool is straightforward to install via Smithery CLI and works immediately in Claude Desktop. No environment variables, no authentication dance. It just connects and runs. The language support is useful if you're working with multilingual content, though it depends on whether the video has captions in the first place. If the video has auto-generated captions only, you get what YouTube gives you, which can be messy. Quirks: it won't fetch transcripts from videos that don't have captions enabled. This sounds obvious, but it's easy to forget when you're throwing links at Claude. The error messages are clear, at least. Also, it's a read-only tool. You can't edit or correct transcripts through the MCP, so if the captions are wrong, you're stuck with them unless you export and fix manually. Who shouldn't bother: if you're already using a paid transcription service like Whisper or Deepgram, this won't add much. It's for people who want quick access to existing YouTube captions without leaving their AI workflow. If you need high accuracy or custom vocabulary, look elsewhere. For everyone else, it's a clean, no-fuss tool that does what it says.