AI Meeting Notetaker for Zoom and Google Meet with Local Transcripts
Our own system (RetailBox)
Retailbox: our own internal system, 2026
A meeting bot for our own team: it joins Zoom and Google Meet calls from the team's calendars, transcribes them with speech-to-text that runs on our own Macs, and files the notes into the right project.
What was the problem?
We wanted every call captured without sending client audio to a cloud notetaker.
How did Retailbox solve it?
A bot joins Zoom and Google Meet calls from the team's calendars (Teams is in beta), one bot per meeting, with a second Mac as backup for Zoom. Speech-to-text runs on our own Macs: live text during the call and a clean transcript after it. A local model writes the summary, then Claude files the notes into the right project and the team's memory graph. Client recap emails are only drafted; nothing goes out on its own, and the admin panel has no send endpoint. The admin panel is reachable only from approved devices on a private network.
Not everything is local: Claude routes the notes after the local steps. What stays on our Macs is the audio and the speech-to-text.
Facing the same problem? These are the services that cover this kind of work: AI agent development services.
What does it look like?
(opens the full-size image in a new tab)What were the results?
- ✓70+ calls transcribed by the bot between October 6 and November 12, 2026
- ✓Live text lags speech by about 1 s on Google Meet and about 3 s on Zoom
- ✓Client audio is not sent to a cloud speech-to-text service
- ✓Client recap emails are drafted for a person to approve, never sent by the bot