UnisonWhat is AI Memory?
AI memory is the layer that lets agents retain and recall knowledge across sessions - distinct from the context window, distinct from RAG. Types, architecture, and how teams share it.
AI memory is the layer that lets an agent retain knowledge across sessions and recall the relevant part when it matters - the capability a context window cannot provide, because the window resets every session and degrades as it fills.
What counts as memory (and what doesn't)
| Layer | Persists? | Shared? | What it is |
|---|---|---|---|
| Context window | no - resets per session | no | working memory |
| RAG index | yes | yes | what you wrote down beforehand (memory vs RAG) |
| AI memory | yes | can be | what agents learned while working |
Memory subdivides into episodic, semantic, and procedural - what happened, what is true, how to do things. Production systems also need supersession (truth changes), abstention (admitting ignorance beats inventing), and - the part most tools skip - sharing: memory scoped to one user dies with that user's sessions.
Personal vs team memory
Most AI memory tools scope memory to a user id. That solves "my assistant remembers me" and nothing else. The harder, more valuable problem is team memory: many agents and humans writing one accumulating corpus with visibility boundaries and conflict handling.
Unison is AI memory built for that case: one API for ingest and recall, a bitemporal knowledge graph underneath, and scope-by-path sharing from private scratch to company truth.
What is a Company Brain?
A company brain is shared organizational memory that AI agents and employees both read and write - decisions, facts, and context that survive personnel and session turnover.
Memory types
Episodic, semantic, and procedural memory for AI agents - what each type stores, when to use which, and how Unison classifies them at search time.