Privacy

Privacy

Private by architecture, not by promise. There's no account and no server of ours. Recording, transcription (including live transcription) and storage all run on your computer.

What can go online, and when

The app connects to the network only when you do something that needs it:

What	When	Where
Speech model download	You click Download / Set up transcription	Hugging Face (ggerganov/whisper.cpp)
Update check	You click Check for updates, or you turned on the launch check	GitHub Releases
GitLab sharing	You click Upload, Test connection or create issues	Your GitLab server
Company calendar	You open Calendar, test, detect or approve an event	Your calendar server
LLM summary	You click Summarize. Ollama stays local; API / Open WebUI / OpenCode / Claude Code send the transcript text to that provider.	The endpoint you configured

Audio is never uploaded. Integrations send text only.

What the app never does

• No analytics, telemetry or crash reports. Not anonymous, not opt-out: none.
• No accounts.
• The built-in whisper.cpp is built without its network features (WHISPER_CURL=OFF, GGML_RPC=OFF, no server).
• The recording engine (cpal + libopus) talks only to your OS sound service (WASAPI, CoreAudio, PulseAudio/PipeWire).
• An automated test (src/privacy.test.ts in the app repository) fails if the code gains a new way to reach the network.

The OS web view (WebView2 / WebKit) follows your operating system's own diagnostic-data settings.

Where secrets are stored

GitLab tokens, calendar passwords and LLM API keys are stored only on this device, in the app's local storage.

The website

The landing page uses cookieless Vercel Web Analytics and reads public release data from GitHub. It stores nothing on your device. See Legal Pages and Analytics and Download Counts.