What gets captured, and where it lands.
A meeting starts. Seriesly's menu-bar daemon detects the Zoom or Google Meet window and asks you, the first time, whether to listen. If you say yes — and only if you say yes, with a one-click skip for any meeting you'd rather it sit out — the audio is captured through macOS's standard audio frameworks and written to a temporary file on your disk.
You are responsible for complying with any applicable recording-consent laws and for notifying the other participants in your meetings where required. Several US states (including California, Florida, Illinois, Maryland, Massachusetts, Pennsylvania, and Washington) require all parties on a call to consent before it can be recorded. Seriesly does not announce itself to other participants; that's your call to make.
A local speech model transcribes that file. The transcript and the recording are written to a SQLite database in your Application Support folder, where your disk encryption is the layer protecting it at rest. Everything after that — the briefs, the open loops, the topic pages — is computed on your machine from that database, sometimes ahead of time so it's ready when you need it.
Your meeting database lives on your disk, not ours; there is no Seriesly server holding transcripts at rest. When the app needs a model to draft a brief or answer a question, the relevant slice — excerpts, names, action items — passes through our proxy to Anthropic, OpenAI, or Google. The proxy logs request traffic on our side for debugging and abuse prevention, and the providers handle the call under their own logging and retention policies (each publishes its own). If that's tighter than you can tolerate, the bring-your-own-key option in settings sends the call directly from your machine to the provider on your account — we do not see it, and nothing about that request lives on Seriesly infrastructure.
Seriesly is for adults: you must be at least 18 to use the app. If you are buying it for a team, the same applies to every seat.