Step 01
Start recording from menubar or CLI
One click in the native macOS menubar app (or command line for power users).
MeetingMind workflow
How the app works, step by step.
This is the real pipeline from recording to usable notes. No team dashboard required — just clean outputs you can search, track, and reuse.
Local-first: recordings and meeting history are stored locally. If you enable AI post-processing,
relevant text may be sent to an external model provider.
End-to-end workflow
Step 02
Capture audio in reliable chunks
Session audio is written in 5-minute WAV chunks to reduce failure blast radius.
Step 03
Transcribe while recording
Whisper transcription runs during the session so processing doesn’t start from zero at the end.
Step 04
Apply speaker labels
pyannote diarization adds speaker structure. If diarization fails, MeetingMind falls back to plain transcript.
Step 05
Generate structured meeting output
AI post-processing produces title, summary, decisions, action items, and participants.
Step 06
Store + search locally
Meetings are stored in SQLite with FTS5 full-text search for fast retrieval.
Step 07
Act on outcomes
Track action items (open/done), relabel speakers, export to markdown, and query via MCP in Claude Desktop.
What you can do today
Recording controls
- Start/stop from menubar
- Record from CLI
- Visible recording status
Meeting memory
- Search all past meetings
- Find decisions fast
- Keep local history
Output control
- Manual speaker relabeling
- Action item state tracking
- Markdown export for Obsidian/LLMs
MeetingMind is built to reduce overhead. You decide when to run it, what to capture, and where the output lives.