Lives in your menu bar. Hears both sides of the call — yours and theirs. Writes the transcript, pulls out the decisions, tracks the commitments, and hands you a clean record when the meeting ends. Audio never leaves your machine unless you choose.
Open source · Bring your own API keys · No accounts, no telemetry, no servers
How it works
Three things happen in the background while you're in the meeting. By the time it ends, you have a record you can act on — not 90 minutes of audio you'll never replay.
CoreAudio captures mic and system loopback in parallel — you on one channel, them on the other. Streamed to Deepgram for realtime diarized transcription.
Claude (or GPT as fallback) reads the running transcript and extracts decisions, open questions, action items, and explicit commitments — with the source sentence attached to each.
The whole meeting lands as a single JSON file in ~/Documents/Nmonic/. Searchable, exportable, yours. No cloud account, no sync, no upload.
What it captures
Speaker-attributed, time-stamped, both channels separated so the host and guest are never confused.
"I'll send it by Thursday." Pulled out with the owner, the due date, and the exact sentence it came from.
What was settled, in one line each — so you can scan a meeting in 30 seconds two months from now.
Things that came up but weren't resolved. The list you actually want before the next sync.
macOS-native
Nmonic isn't a webapp wearing a chrome. It's a native SwiftUI menu bar app that wakes up when you start a call, captures the audio Apple normally hides, and goes quiet the second the meeting ends.