What MCP is
MCP (Model Context Protocol) is an open standard for connecting AI assistants to tools and data sources. Instead of copy-pasting a transcript into a chat, an MCP-capable agent calls tools directly. MeetConnect ships a local MCP bridge, so the connection stays on your machine.
The tools the bridge exposes
The bridge serves the live caption store over six read tools:
| Tool | What it returns |
|---|---|
| meeting_status | Whether a meeting is live and basic state |
| get_recent | The most recent transcript turns |
| get_full_transcript | The full coalesced transcript |
| search_transcript | Matching lines for a query |
| list_speakers | The speakers seen in the meeting |
| list_meetings | Meetings known to the bridge |
What you can ask
- “What did we just decide about pricing?” — mid-call, from the live transcript.
- “Who raised the concern about the deadline?” — via speaker + search.
- “Summarize the last ten minutes.” — from the recent turns.
What needs the bridge (and what doesn't)
The MCP bridge and AI transcription are the bridge-connected mode — a small local Node process you run with your own key. Free live caption capture, .txt export, and browser-only audio recording all work without it. See recording Meet locally.
NoteToday the MCP tools serve the live, in-memory caption store. A durable, archive-backed index for 'talk to all my past meetings' is on the roadmap, not shipped — we keep that distinction honest.