Scribiz

Scribiz gives Claude the context of a video from a link. Claude can start from a short overview with chapters and key moments, search the video for a word or phrase, read just the part it needs, or ask a question and get an answer with 3 to 5 cited moments. Every moment comes with a link that opens the video at that time, so you can check the answer yourself.

Five tools, all read-only:

• Get video context: title, length, summary, chapters and key moments.
• Search a video: where a word or phrase is said, ranked, with links.
• Ask a question about a video: an answer with cited moments.
• Get a video transcript: the words, one page at a time, as TXT, Markdown, SRT, VTT or JSON.
• Check a video job: picks up a video that is still processing.

In this connector, Scribiz uses a video's captions when it can read them, serves videos it has already processed, and otherwise has a model read the link. That runs inside a free daily allowance of about 10 minutes of reading, for videos up to 15 minutes. Times are approximate when a model read the link, there are no speaker labels, and Scribiz never looks at the picture, so there are no on-screen notes. When the allowance is used up, captions and processed videos still work, and the error says when it resets.

It takes public links only. It cannot read files on your computer.

Text from a video can say anything. Scribiz wraps it in marked untrusted blocks and flags every result, so Claude can tell it from instructions.

Documentation: https://scribiz.com/docs/mcp