The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Capslane listing page.
The Capslane MCP server retrieves timestamped YouTube transcripts from compatible assistants and developer tools. It exposes three read-only tools for requesting a transcript, checking a generated transcript job and listing the caption languages observed for a video.
https://capslane.com/mcpAuthorization: Bearer YOUR_API_KEYKeep the API key in the client's secure connection settings. Do not include it in prompts.
Node.js 20 or later is required.
get_youtube_transcriptRetrieve existing YouTube captions or start and optionally wait for a generated transcript.
Inputs:
url: public YouTube URL or 11-character video IDlang: optional preferred language codemode: native, auto or generatetext: return one text string instead of timestamped segmentswaitForCompletion: wait for an accepted transcript job before returningget_transcript_statusReturn the current state of a generated transcript job.
Input:
jobId: job identifier returned by get_youtube_transcriptlist_available_languagesReturn the caption languages observed for a public YouTube video. This consumes one transcript request.
Input:
url: public YouTube URL or 11-character video IDnative returns existing captions and never starts speech transcription.auto uses existing captions first and generates a transcript only when needed.generate creates a transcript from the video audio.Create a dedicated API key for each MCP client. Keep keys out of prompts, committed files and shared configuration.
MIT