A client changes the scope. Someone commits to a deadline. The call ends, and the details start to fade. We built Rebrief to preserve the decisions, commitments and reasoning that move work forward. The app turns conversations into a searchable record, with the core AI running entirely on the Mac.
Visit rebrief.coRebrief records meetings across Zoom, Teams, Google Meet and other calling apps without a bot joining the conversation. Automatic meeting detection reduces the manual work around each call. The result is a consistent way to capture meetings, regardless of which platform a client or colleague uses.
The useful part of a meeting is what happens because of it. Rebrief turns discussion into decisions, action items and unresolved questions. Consultants gain material for a client brief. Product and engineering leads retain the reasoning behind a change in direction. The goal is a record that supports the next piece of work without another round of note-taking.
A commitment loses meaning when it is detached from the person who made it. Rebrief's optional speaker identification connects contributions to the people behind them. Voice recognition carries that context across meetings, helping researchers preserve interview attribution and project leads follow commitments over time.
Questions rarely arrive in the same order as meetings. A client commitment from last week may matter to a decision today. Rebrief retrieves relevant material from a local meeting index before generating an answer, grounding each response in recorded conversations. This makes the meeting archive useful across projects and over time, beyond the summary of any single call.
Recording, transcription, notes and local answers run on the device. After the required models download, the core workflow works offline. Speech recognition supports 99 languages with automatic language detection. Meeting content stays on the Mac.
Meeting records need to travel into briefs, research and follow-up work. Open export formats let that information move into other tools. An optional, read-only Model Context Protocol (MCP) connection extends access to external assistants. The local workflow remains independent of those services, with sharing controlled by the user and any shared content subject to the receiving tool's data practices.
Native macOS engineering brings recording, transcription and local AI into one application. The entire meeting workflow runs on the Mac, without a cloud processing service.
On-device speech recognition turns meeting recordings into searchable transcripts. Background processing allows the next meeting to be recorded while the previous one is still being transcribed.
Local AI turns meeting content into summaries and answers grounded in recorded conversations. Models are selected to suit the Mac's available memory.
Unfinished transcription resumes after relaunch, and a failed job does not block the next meeting. Recording files preserve audio already written if the app stops unexpectedly.
Keeping meeting content on the Mac shaped decisions across the stack. Models compete for memory with other applications, and processing must survive interruptions. Model selection, background queues and recovery are as much a part of delivering useful local AI as the quality of its answers.
A polished summary can still invent a commitment. Our evaluation criteria check coverage of the whole meeting, factual support and whether actions reflect what people actually agreed to do. The same standard applies to answers across meetings: claims must stay within the retrieved evidence. That standard guides changes to prompts and models.
The value of a meeting record often arrives days after the call. Someone needs the reason, the commitment or the exact words. Rebrief connects capture, notes and retrieval in one shipped app, so a conversation can keep informing the work that follows.