Meeting Assistant has no account and no server of ours. Audio is recognized on the device, your documents are indexed locally, and the only text that ever leaves is what you send to the model you chose.
Direct download · version 0.1.0 · 3.6 MB · macOS 26 or later, Apple silicon · notarized by Apple.
How it helps
Both audio streams are recognized by Apple’s engine or a Whisper model running on the Mac. The audio is not stored by the app beyond what recognition needs.
Your files are split and indexed in a local database with on-device embeddings. Meetings are JSON files in your Library. Nothing syncs anywhere.
With Apple’s on-device model, no text leaves the Mac. With ChatGPT, an API or Claude, the recent transcript and matching passages go to that provider under its terms, and only when an answer or notes are made.
There is no account and no analytics inside the app. What a model provider receives is governed by the provider you pick.
Yes, with Apple’s speech engine or a downloaded Whisper model, and Apple’s on-device model for notes and answers.
As JSON files in ~/Library/Application Support/MeetingAssistant/Meetings/ on your Mac.
Often you must. Many places require consent from everyone on the call.
