AI-native meeting capture

Don't lose
the meeting.

MeetConnect taps Google Meet's caption stream directly — not the screen — records the audio crash-safe, and turns every conversation into a clean, speaker-attributed record your AI agents can read live.

Runs on your machine Bring your own key Free caption export

live demo · captions + audio + an agent answering mid-call

~95%
capture engine done
≤10s
lost on a crash — that's it
2 langs
Persian + English, as spoken
6 tools
agents call over MCP
The mechanism

Most tools read the screen. We tap the signal.

Google Meet streams captions over a WebRTC data channel. MeetConnect decodes that channel at the source — so the transcript is the real thing, not a guess scraped off the DOM.

01

Tap the channel

An injected hook decodes the caption protobuf off Meet's data channel and the speaker roster off the collections channel — at document-start, before the page even paints.

02

Stitch & record

Segments are de-duplicated and upserted into a clean transcript while audio records in crash-safe ~10-second chunks. Refresh, pause, take a break — it resumes.

03

Connect to AI

A local bridge serves the transcript over MCP. Your agent reads the meeting live, searches the archive, and — with Gemini — produces mixed-language transcripts from the audio.

Meet tabinline hookservice workerws://127.0.0.1local bridgeMCP · agent
What's in the box

The hard, valuable engine — already built.

The capture core is done and dependable. Here's what it does today.

Live caption capture

The real Meet caption stream, decoded at the source — draggable live panel, auto-enabled captions, instant .txt export. Free, forever.

Crash-safe recording

All participants mixed to one file, chunked every ~10s to disk. Pause and resume, recover orphaned audio, finalize to a single seekable .webm.

Gemini transcription

Optional, never forced. High-quality verbatim mixed-language transcripts, an 'English for AI' mode, and incremental 'new part' passes mid-call.

MCP bridge

One local node process exposes six tools — status, recent, full transcript, search, speakers, meetings — so any MCP agent reads the room live.

Speaker attribution

Names resolved from the collections channel and DOM correlation, de-duplicated as Meet bumps caption versions. Falls back gracefully.

Yours by default

Capture, record, and transcribe entirely on your own computer with your own API key. Nothing leaves the machine unless you decide it should.

Privacy-first, not as an afterthought

Your meetings are the most candid thing you say all day. Keep them.

Cloud-only tools make a copy of every word you speak. MeetConnect is built local-first: the whole pipeline can run offline, on your hardware, with your key. The free base needs no account and no upload. When you choose the cloud, you choose it on purpose — for the dashboard and the heavier AI, never as the price of admission.

No account to start
No meeting leaves by default
Bring your own API key
The north star

Talk to your meetings.

Every conversation becomes a searchable, semantic record. Ask what was decided, who raised a concern, what you committed to last Tuesday — and get the exact moment back, with a reference to the source. Live during the call, and across the whole archive after.

you
what did we commit to on the landing site?
meetconnect
Ship it this week — Meghdad, 00:21. Capture engine flagged as done.
you
any open concern about the cloud tier?
meetconnect
Pricing only. $20/mo agreed, no objections logged.
Pricing

Free where it should be. Paid where it's worth it.

The live capture and export are free because Google gives us the captions. You pay only for the heavy lifting — and even then, you can bring your own key and stay local.

Free
$0forever

The on-ramp.

  • Live Google Meet caption capture
  • Instant .txt transcript export
  • Crash-safe audio recording
  • No account, nothing uploaded
Add to Chrome
Most private
Local
$5lifetime

Offline, yours, bring your own key.

  • Everything in Free
  • Gemini AI transcription (your key)
  • Mixed-language verbatim mode
  • Local MCP bridge for agents
  • Lifetime updates
Get the license
Cloud
$20/ month

We do the heavy lifting.

  • Everything in Local
  • Managed AI — no key needed
  • Cloud dashboard & archive
  • Meeting summaries & categorization
  • Talk-time analytics
Join the waitlist

Pricing is a launch sketch — privacy-first stays free. Final tiers may shift as we learn.

Questions

The honest answers.

How is this different from screen-scraping tools?

MeetConnect taps Google Meet's WebRTC caption data channel directly instead of reading the DOM. The live transcript is the actual caption stream — faster, cleaner, and speaker-attributed, the same approach Tactiq uses.

Is my meeting data private?

Yes. It's privacy-first and local-first. Capture, recording, and transcription can run entirely on your own machine with your own API key. The free base needs no account and uploads nothing.

Can an AI agent read the meeting while it's happening?

Yes. A local MCP bridge exposes the live transcript over six tools, so an agent can answer 'what did we just decide?' during the call and search past meetings afterward.

Does it handle mixed Persian and English?

Yes. With Gemini it produces high-quality mixed-language transcripts, keeping each language as spoken, plus an optional clean-English mode for AI workflows.

Do I have to use the AI transcription?

No. AI transcription is always optional. The free tier — live captions, export, and audio recording — works fully on its own.

What does it cost?

Live capture and export are free. A local lifetime license (~$5, bring your own key) unlocks AI transcription and the MCP bridge; a cloud tier (~$20/mo) adds managed AI, a dashboard, and analytics.

Capture your next meeting like it mattered.

Add MeetConnect to Chrome, join a Meet, and watch the transcript stream. Free to start — no account, nothing to lose.