How Hathi remembers everything (without being creepy)

Memory is what turns a chatbot into an assistant. Without it, every message is a cold start. With it, Hathi becomes someone who knows you — your habits, your people, your plans.

From messages to meaning

When you message Hathi, it doesn't just read your text — it understands what matters. Key facts are extracted and distilled into structured memories: "Sarah's birthday is March 12th," "You prefer aisle seats," "The villa booking is under the name Patel."

These aren't stored as raw transcripts. Hathi breaks them into discrete, useful facts stored in your private database table. Each memory is tagged with an icon and a source, so you can see exactly what Hathi knows and where it learned it.

Your memory dashboard

Every user gets a personal Memory page. You can browse everything Hathi remembers, delete individual facts, or export the entire collection as a JSON file. Nothing is hidden. Nothing is locked away.

This transparency is by design. If an assistant is going to remember your life, you should be able to audit and edit that memory at any time.

"Hathi treats memory as a shared resource — something you both build and both control."

Security by default

Your memories live in a PostgreSQL database protected by Row Level Security (RLS). That means each row is gated by your user ID — even if someone accessed the database directly, they couldn't read your memories without your credentials.

Hathi also cleans up responsibly. If you pause your subscription, your memory is retained for 90 days so you can resume seamlessly. If you delete your account, everything is erased within 30 days. You remain in control, always.