Every team has the person who "knows where that talk is". Maybe it's you. The research your company runs on — conference talks, competitor launches, expert interviews, your own webinars — lives in watch-later lists, dead Slack links and someone's memory. When a decision needs the source, the ritual begins: who watched it, which video was it, twenty minutes of scrubbing for one quote.

A link dump is write-only memory. What you actually want is a curated library that answers questions: one place where the channels that matter are imported, transcripts and all, and where "didn't their CTO explain exactly this on stage?" takes seconds to answer — with a citation to the moment, not a vibe.

library_import_youtube_channel does the curating: point it at a channel or playlist — a conference archive, a competitor's channel, your own webinar backlog — and it imports every video into your veridive library, transcripts included. library_search and its sibling library_list are your card catalog: find items by title, tag or type in an instant, free.

deep_context is where the library earns its keep. Pick the imported talks that matter — your assistant scopes a selection by their library IDs — and ask your real question across them. It runs the full research pipeline over those transcripts and returns a grounded answer where every claim cites the exact chunk of the exact talk, so nobody has to rewatch anything to trust it.

And because the connection is just an MCP config, the library isn't locked to your machine: share the connector setup the way your team already shares tool configs — a project-level mcp.json in the repo, a shared Claude connector — and the same curated brain answers inside everyone's assistant.