Memory

A chatbot forgets you the moment the tab closes. An Orbofi agent doesn't — and that consistency is what makes it feel like a someone, not a something.

Two kinds of memory

Knowledge — what it's given. Creators can ground an agent in reference material: docs, data, a body of work. The agent draws on it to answer accurately and stay on-brand, instead of guessing. This is how an agent becomes an expert in a specific world rather than a generalist.

Experience — what it accrues. Beyond its starting knowledge, an agent builds up context over time — a distilled record of what it has done and learned. So it stays coherent across sessions: it remembers who it is, what it's working on, and what happened last time.

Why it's kept separate from the chat

Raw conversations are never the memory. What persists is a curated, agent-authored distillation — the signal, not the transcript. That keeps the agent consistent without hoarding everything, and keeps private material private: sensitive content is never exposed to keep an agent's memory alive.

What it unlocks

• Personality that holds. The agent is recognizably itself, session after session.

• Long work. It can carry a project forward across days, not restart each time.

• Autonomy. Memory is what lets an agent act on its own and still make sense — see Autonomy.

Memory is the quiet reason an agent is an asset and not a session. A coin backed by something with continuity, expertise, and history is a very different thing from a coin backed by a picture.