Privacy
What we collect, where it goes.
Short version
Purr runs entirely on your Mac. We do not collect anything. We have no servers, no analytics, no account system, and no way to see what you say into the app. There is nothing for us to leak because there is nothing for us to store.
What the app does on your machine
When you hold the hotkey, the app records audio through your microphone, runs speech recognition on the Apple Neural Engine, and types the result into whichever app has focus. The audio is held in memory only for as long as it takes to transcribe. We never write recordings to disk and we never send them anywhere.
Meeting mode saves the transcript (text, not audio) to
~/Documents/Purr Meetings/. Those files live on your
machine. We do not see them.
The only network requests Purr makes
- Speech model downloads. On first use the Parakeet TDT v3 model (~600 MB) downloads from Hugging Face. If you switch engines, Whisper variants (39 MB to 1.5 GB) download from the same place.
- Optional summarizer. If you turn on meeting summarization, the Gemma 3 4B model (~2 GB) downloads once from Hugging Face.
After those downloads finish, Purr makes zero outbound requests. You can verify with Little Snitch or the macOS firewall.
What this website collects
Nothing. There are no cookies, no analytics scripts, no third-party embeds. The site is plain HTML served by Netlify; their access logs may record IP addresses for abuse prevention, governed by Netlify’s privacy policy.
Children
Purr is not directed at children under 13. We do not knowingly collect data from anyone, so there is no separate flow for minors.
Changes
If this policy ever changes, the updated date at the top of this page will move forward. The full history is in the git log of this file.
Contact
Questions? Open an issue on GitHub.