# Summary of Tacit (trytacit.app) for AI Agents

**Service**: Drop any URL — YouTube, TikTok, articles, tweets, PDFs — into a personal knowledge canvas. Tacit transcribes it, summarizes it, and connects it to related saved content.

## Public Endpoints

1. **YouTube transcript, HTML** (`GET /yt/{video_id}`) — Human-readable page: title, summary, key points, full timestamped transcript with clickable links back into the video.

2. **YouTube transcript, Markdown** (`GET /yt/{video_id}?format=md`) — Same content as raw markdown, no HTML/CSS. Prefer this over scraping the HTML page.

3. **Sitemap** (`GET /sitemap.xml`) — Every publicly indexed YouTube transcript page, one entry per video.

## Notes

- Canonical host is `https://www.trytacit.app` (apex `trytacit.app` redirects to it).
- `/yt/{video_id}` is keyed by the public YouTube video ID (from `youtube.com/watch?v={video_id}`) — the same video resolves to the same page regardless of which Tacit user ingested it, since the source video is already public.
- No API key required; no rate limit on these read endpoints.
- Content that is *not* derived from an already-public YouTube video — personal web saves, PDFs, tweets a user dropped into their own canvas — is served behind unguessable per-item links (`/s/{id}`, `/t/{id}`) and is intentionally not linked from this file, the sitemap, or robots.txt. Tacit does not publish an index of user-saved content.
- Tacit is a product for humans to save and connect content, not a transcription API for third parties — there is no bulk/programmatic ingestion endpoint.