> For the complete documentation index, see [llms.txt](https://whitepaper.huralya.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://whitepaper.huralya.com/the-intelligence-behind-hura.md).

# The intelligence behind Hura

Hura's AI is designed to feel like a thoughtful person, not a chatbot. Three ideas make it different.

**A coach that leads.** A Hura session has a shape. The coach checks in, explores what is going on, works on one useful insight or skill, then closes with a short recap and one small next step. People leave a session with something in hand, not an endless thread.

**A memory that grows.** Every conversation, journal entry, test and journey teaches Hura a little more: what matters to the person, what triggers them, what helps, who is in their life and how they are changing. Every coach and companion draws on this one shared memory, so the user is known across the whole app. Old details fade naturally, and deleting the account erases everything.

**Efficiency by design.** Hura sends the AI only what it needs: a compact profile, a running summary and the latest turns. This keeps replies fast and the cost per user very low, which is what makes a generous free experience possible at global scale.

Hura is model-agnostic. It chooses the best model for each task, a strong empathic model for conversation and lighter ones for background work, and can adopt better models as they appear without changing the app.
