Your memories should belong to you.
Cappy is designed to work locally by default and to make cloud features a visible choice. This page explains the current product behavior; it is not a promise that every external model provider follows the same rules.
Open your data controls →Data kept on your device
Chats, saved memories, tasks, preferences, and optional media attachments are stored in browser storage. The on-device model and local transcription assets are cached by the browser. You can export a JSON backup, import it later, clear model caches, or delete all local data from Privacy & data.
When data leaves your device
- If you choose an API model, chat prompts and relevant memory context are sent directly to that provider.
- If cloud memory understanding is enabled, a captured note or image may be sent to the selected provider for structuring.
- If live web search is enabled and a question needs current information, the query is sent through the cappy.pet search service to TinyFish, Exa, or Firecrawl, depending on your selected source. Wikipedia searches are sent directly to Wikimedia.
- If you request weather, coordinates or a city query are sent to the weather service.
Website analytics
The production website loads Microsoft Clarity by default to collect page visits, device and browser information, clicks, scrolling, and masked session recordings so we can improve Cappy. All page content is marked for masking before the recorder loads, including chats, memories, tasks, and forms.
Clarity runs in cookieless mode: analytics and advertising storage permissions are denied, and no analytics consent prompt is shown. Clarity may still receive usage data in this mode. Local development, preview hosts, and the desktop app do not load Clarity. See Microsoft Clarity’s privacy information for how Microsoft processes analytics data.
Credentials
On the macOS desktop app, model-provider keys use the operating system’s protected storage when available. On the website, model and optional search-provider keys remain in this browser’s local storage. The free TinyFish search option does not require your own key. If you select Exa, TinyFish API, or Firecrawl, your key is sent through the cappy.pet search service only for that request and is not stored there. Do not add a key on a shared or untrusted device.
Your controls
Choose explicit-only, suggested, or automatic memory. Disable cloud memory organization. Edit or forget individual memories. Export a backup before clearing the browser. Delete all local data at any time.
Last updated: October 3, 2026.