Team Meet turns live conversation into sourced notes, decisions, action items, diagrams and share-ready clips — before anyone leaves the room.
Team Meet does not bolt a note-taker onto somebody else’s call. The intelligence lives in the room, so everything can happen while the conversation is still alive.
Decisions, open questions and the real points of disagreement.
00:38
No prompt. The brief is the conversation itself.
Start simple. Switch when the meeting reveals what it actually needs. The new mode re-reads everything already said.
“The room is the platform. The API proves it.”
Jack · 18:42Raised at 15:09 · No clear answer yet
Due Friday · Confirmed out loud
JackCommitMayaCommitNoahSupportLenaCommitThe browser composites every tile into one 1080p frame and uploads while you talk. Stop recording, and the share link is already there.
See recording flow ↗
Call on the AI out loud. It keeps what this meeting established separate from outside knowledge — and drops a claim when it cannot find the quote.
Meet Ova ↗You decided to ship the public Agent API first, with Maya owning the quickstart.
“Let’s ship the agent API first…”Jack · 18:42 · Jump to quote
A summary is only useful when you can trust it. Team Meet ties every note to the exact words, speaker and moment it came from.
Speaker identity is structural. Each microphone arrives on its own connection — no “Speaker 1,” no guessing.
Quotes stay attached. Click any claim to hear the room establish it.
Unproven claims get dropped. The AI does not dress uncertainty up as fact.
Let’s ship the agent API first, then open the room to every workflow.
I’ll own the quickstart and have it live by Friday.
I support that order. Webhooks can follow in the same release.
Your bot joins with one socket and a key — not a headless browser pretending to be a person.
Live room
Ground-truth transcript
Start a meeting, read every line with a stable speaker identity, speak into the room, draw, post notes and receive the finished recording.
const ws = new WebSocket('wss://teammeet.io');
ws.on('open', () => ws.send(JSON.stringify({
type: 'agent-hello',
key: process.env.TEAMMEET_KEY,
meeting: 'abc-defg-hij'
})));
ws.on('message', raw => {
const m = JSON.parse(raw);
if (m.type === 'transcript') {
console.log(`${m.line.speaker}: ${m.line.text}`);
}
});A name, a click and an unguessable link. That is the whole setup.
This meeting needs the host's approval. You will drop straight in when they say yes.
These are yours alone. Nobody else in the meeting is affected.
Your microphone is transcribed on your own machine, so this is the language you are recognised in — everyone else keeps theirs.
Your screen is the only lamp most people have pointed at their face. In an unlit room a dark interface wastes it. These are yours alone — nobody else sees them, and neither does a recording.
Only the host can change this.
Only the host can change this.
Building one? The agent API is open and free.
–