On Zoom, on Meet, or in person. Gijiroku records the sound playing on your Mac and your own voice, then turns it into a transcript and minutes — automatically. Your audio never leaves this Mac.
macOS 26 or later ・ Apple Silicon
Not flashy features — just what minutes actually need, done reliably.
Zoom, Meet, Teams — and in-person or one-on-one talks. Whatever the app, it records the sound on your Mac and your own voice together. Always the same way.
Turning speech into text happens only on this Mac. Neither the recording nor the text is sent anywhere.
The result is plain Markdown. Drop the folder into iCloud or Obsidian and read it from your iPhone. Not minutes that vanish when a subscription ends.
All you do is record the meeting. The rest is three steps.
Only the summary step sends anything out — and only the transcript text, to an AI you choose (Claude, Codex, and more). You can also skip summaries entirely.
Meetings come back to back — so nothing here blocks your next one.
Minutes are written in the background. Even with back-to-back meetings, there is no waiting.
Progress like "Organizing 3/8" and elapsed time are always visible — no wondering whether a long meeting got stuck.
Names and companies you register in the dictionary are written exactly as registered — unknown names never get a guessed spelling. Per-folder dictionaries included.
Summarize with Claude or Codex — or connect any provider with an OpenAI-compatible API key.
Below is an example of the actual minutes layout (a sample, fictional meeting).
Each timestamp (00:31:10) links back to that moment in the transcript. “Who said what” is something you can check against the record.
What leaves your Mac and what doesn’t. It’s settled by design.
| Never leaves your Mac | Audio ・ recording files ・ the full transcript. None of it leaves this Mac. |
|---|---|
| What does leave | Only the transcript text used to write a summary. It’s handed to the AI you chose (Claude, Codex, or another provider) — never to one you didn’t pick. |
| With summaries off | Nothing leaves at all. Everything up through transcription finishes on this Mac. |
“Doesn’t transcription use AI too?” — Yes. But that AI runs on speech-recognition models installed on your Mac (no network, processed right there). The only time anything goes to an outside AI is for the summary. So with summaries off, nothing leaves.
Please record only with participants’ knowledge. The app also includes a way to announce that recording is on.
A beta build is available now, for macOS 26 or later on Apple Silicon.
July 10, 2026 ・ DMG 2.5MB ・ macOS 26+ / Apple Silicon
(Check yours: Apple menu → "About This Mac" — an "Apple M…" chip means Apple Silicon)
Download password (included in the invite you received)
SHA-256: 7bc6f0350d0d87a0019c30dccfdf0a472c858b00f0998a78cf423cd0ea4b0040
This beta is not yet developer-signed. On first launch, macOS will say it "could not verify this app is free of malware". When that happens:
We’ll email you when the Windows version is ready.
The signup endpoint isn’t connected yet. Your address will be used only to notify you about the Windows version.