A straightforward wrapper around YouTube's transcript API and yt-dlp that exposes five tools for pulling video content into Claude. You get get_transcript with multiple format options (text, JSON, SRT, WebVTT), get_video_metadata for channel and chapter data, list_playlist_videos for batch enumeration, and summarize_transcript which structures the output so you can review the prompt before execution. Handles multi-language transcripts with fallback priority and parses any YouTube URL format. Particularly useful when you need to analyze video content without watching it or want to build workflows around playlists. The metadata blocks are cleanly separated and optional, so you can strip them out when you just need raw transcript text.
claude mcp add --transport stdio zlatkoc-youtube-summarize uvx youtube-summarize