Record on purpose.
Most AI recorders listen all the time and hope nobody minds. Scribe records only when you tap, tells everyone it's recording while it does, and keeps the audio on your phone. What you get back is not a summary. It is a debrief: every claim extracted, scored, and given a verdict with evidence.
Join the waitlist
What makes Scribe different
Consent is code, not a promise.
Scribe has no always-on mode to turn off. A recording begins when you tap record and ends when you tap stop; nothing listens in between.
While it records, the screen says so and the app announces it out loud at the start, so the people in the room learn it from the room rather than from a policy page.
The audio stays on your phone by default.
A recording is written to your device and nowhere else. Transcription runs on the phone, so the default path uploads no audio at all.
You can turn on cloud transcription for a recording that needs it. That is an explicit, per-account choice, it is metered, and it is the only way audio leaves the device — which is what the words "by default" are doing in the sentence above.
You get a debrief, not a summary.
A summary tells you what was said. A debrief tells you what was claimed: each claim is pulled out, scored for confidence, and given a verdict — verified true, verified false, unverifiable, or not yet checked.
Every verdict names the utterance it came from, and the source it was checked against when there was one. A claim with no source keeps no source rather than borrowing one.
Read further
- See a debrief
One worked example, end to end: six claims, four verdicts, three things to go and do.
- The consent report
Who has to agree before you press record, in all fifty states and the District of Columbia.
- Against the alternatives
Pocket, Plaud and Otter — including the rows where they win and the one nobody wins.