Summary
A video link stops being a dead end in the middle of a conversation.
Paste the URL and the model has the transcript to work from — summarising a talk, locating the passage where a claim is made, checking whether a tutorial actually covers your case. What it reads is the subtitle track rather than the audio or the picture, so a video published without subtitles gives it nothing to work with.
What it is
A single-tool server that fetches the subtitles of a YouTube video from its URL.
What you get
- The subtitle track of a YouTube video, retrieved from the video URL and returned as text the model reads in the same turn
Requirements
A Node runtime and the URL of a video that carries subtitles. No account and no key. MIT-licensed.
Setup effort
One command — npx -y @anaisbetts/mcp-youtube