Skip to main content
Pass a YouTube video URL to GET /web/scrape/markdown. Context.dev returns the video’s title, channel, duration, view count, description, and transcript as one Markdown document. You do not need a separate transcript endpoint. 1 credit per call

Supported YouTube URLs

Context.dev recognizes:
  • Watch and share URLs, including youtube.com/watch?v=... and youtu.be/...
  • Shorts and live video URLs
  • Embed URLs, including privacy-enhanced youtube-nocookie.com embeds
The video must have captions available. Videos without captions still return the available metadata and description, but omit the transcript section.

Get the transcript

Export your API key, then pass the YouTube URL as the url query parameter:
Install an SDK from the Quickstart, or use cURL directly.
Read the markdown field from the JSON response. Its structure looks like this:
Transcript paragraphs begin with a source timestamp roughly every 30 seconds, so an agent can summarize the video while preserving references to the original moment.

Use the transcript

The returned Markdown is ready to:
  • Summarize with an LLM while retaining timestamp references
  • Index in a RAG or semantic-search pipeline
  • Search for claims, topics, or quotes within a video
  • Create notes, chapters, or searchable video archives
For the complete request and response schema, see the GET /web/scrape/markdown API reference.