Titles, summaries, chapters, and a written document from a recording. Included in paid plans with no API keys to manage, and never required to capture anything.
Steps, screenshots, click highlights, and sensitive-field blurring are all deterministic, derived from real browser events. You can run Knovvy end to end and never touch AI.
That is deliberate. If a generated summary is a bit off you notice, because you are reading it. If step four of an SOP is subtly wrong, the person following it six months from now has no way to know, because they are following it precisely because they do not know the process. Structure has to be deterministic.
Every one of these is a place where you are already reviewing the output, which is where probabilistic is genuinely useful.
A free account gets three one-time AI trial passes, so you can see what a generated summary or a document-from-recording actually produces before deciding it is worth paying for. They are consumed as you use them and do not renew.
Running out changes nothing about capture. A free account with zero passes left still records guides and videos exactly as before.
Using a generative feature sends the relevant text to a third-party model provider. Capture does not. The full breakdown of what leaves our infrastructure and what does not is on the security page.
Transcription is the exception worth naming: it runs on hardware we operate, so audio is never handed to a speech vendor.