The AI behind Kata
Kata’s replies are written by Anthropic’s Claude models, through Kata’s own service. Kata sends only what an answer needs and keeps nothing.
With each message, your device adds the context the answer needs: your goals, recent logs and memory. It travels encrypted, is used to write the reply, and is not kept. Our AI provider does not train on it and deletes it within 30 days.
Settings, Privacy and data, What Kata knows shows exactly what can travel with a message.
Kata picks up your shorthand, your rhythm and what works for you, and keeps short facts in Memory. You set the tone in Settings, Coach.
Your device works out a short, abstract sketch of how you practice: an archetype and a few coarse traits. It helps us improve Kata and can’t be traced back to anything you logged.
- Kata says when it doesn’t know, and never makes up a number.
- It never saves without a card. See Saving what Kata suggests.
- It stays on health and your day.