Privacy by default, trust by design
The durable asset is your memory — not your raw recordings. We delete audio and ground every answer in its source.
The audio lifecycle
Raw media lives only in a temporary buffer and is auto-deleted after a short retention window (default 24–48h).
Capture
Audio uploaded to a temporary buffer.
Transcribe
Multilingual ASR converts speech to text.
Extract
Memory objects are created and embedded.
Delete
Raw audio is purged; deletion is logged.
Text-note bodies persist because they are the source text. Only transcripts, note bodies, and derived memory are kept.
Ephemeral audio
Recordings are deleted after a short retention window. Re-transcription is only possible while media is retained.
Source-grounded answers
Every answer cites its source entry and declines when the evidence is insufficient — no hallucinations.
Workspace isolation
Every object is scoped to a workspace; queries filter by membership, and storage keys are never used as authorization.
Authenticated access
JWT-based auth; every request except register/login is authenticated, with reasonable token expiry.
Secrets outside source
Credentials and provider keys are kept out of source code and configured per environment.
Observable & logged
Deletions are logged and surfaced; the processing pipeline is observable with retryable failures.
Compliance roadmap
A compliance roadmap (workspace roles, object-level sharing, audit logs, retention policies) is tracked and will grow with the product. It does not gate the closed alpha.
Never lose a decision again
Join founders and teams capturing meetings and quick notes in their own language. Useful after your very first entry.