qwwiwi/public-gbrain-agentos/transcript

Transcript

Fetch YouTube video transcripts via Loore (REST, CLI and MCP server). Docs: docs.loore.ai.

YouTube rule: YouTube data (transcripts, metadata, comments, search) comes only from Loore (yt_* tools) or the official YouTube Data API v3. yt-dlp or scraping from your own IP is allowed only for downloading video files, never for data.

Setup

Create an API key in the Loore dashboard (Settings → API and MCP) at loore.ai and export it:

export LOORE_API_KEY=<your-key>

One key works for both REST and MCP. MCP endpoint: https://mcp.loore.ai/mcp.

POST https://mcp.loore.ai/v1/yt-transcript

curl -s -X POST https://mcp.loore.ai/v1/yt-transcript \
  -H "Authorization: Bearer $LOORE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "VIDEO_URL", "include_timestamps": true}'

The answer comes in one envelope: { "status": "ok", "result": { ... } }. Errors use the same envelope: { "status": "error", "error": "...", "details": { } }. Long transcripts are paged (up to 40000 characters per answer); continue with the cursor/offset from the answer.

Related yt_* tools: yt_search, yt_channel, yt_channel_videos, yt_video (lists available captionLanguages), yt_comments.

Errors

Tips

• For long videos, summarize key points first, offer full transcript on request.
• Use include_timestamps: true when you need precise timestamps for quoting specific moments.
• Use plain text (default) for clean text suitable for translation or analysis.
• Check available caption languages with yt_video before asking for a specific language.