Memory lets the Enterprise Search agent learn from past conversations, exactly the way any AI agent does: "Configure how your agent learns from past conversations." Once enabled, the agent can carry facts, events, and learned procedures forward instead of starting fresh every session.
What is the Memory section?
You configure memory with memory profiles — each one defines a kind of memory — and memory sets, reusable bundles of profiles you can share across agents. Memory is off by default, so the agent only remembers once you turn it on.
How do I turn on memory?
Use the Enable Memory toggle — "Activate memory storing capabilities for your AI agent." It's off by default; the memory profiles and memory sets sections only appear once it's on. Enable it if you want Enterprise Search to recall a user's prior questions or stated preferences across sessions, and leave it off if you'd rather every search start stateless.
How do I add a memory profile?
Use Add Memory Profile to define one kind of memory:
| Setting | Options / purpose |
| Memory Type | Semantic (factual knowledge), Episodic (event history), Procedural (skills and rules) — pick the types to retain. |
| Scope | User Specific (private per user), Team Specific (shared within a team), or Agent Specific (shared across everyone using the agent). A platform-provided, read-only Project-scoped profile can also appear. |
| Generation Settings | The model and instructions that decide what's worth remembering, plus the embedding model and refresh frequency. |
| Storage Settings | An optional expiration period so memories age out. |
Clear instructions on the profile matter — they tell the agent what to store and what to ignore.
Each profile appears as a card; its ⋯ menu offers View Memories (see what's actually been remembered), Edit, and Delete (asks for confirmation, since it can't be undone). The menu is hidden when the section is read-only.
How do memories reach the agent's answers?
Retrieval happens at the start of each request, before the model is called:
The platform resolves which memory profiles apply — the agent's own profiles and attached sets, sorted by scope — skipping anything the user opted out of and any team profile they aren't a member of.
The user's message is embedded and matched against stored memories by meaning, pulling a shortlist of candidates.
A strict relevance filter trims the shortlist, and only the survivors are injected into the agent's system prompt as prior context.
The agent never sees the whole memory store, only a few memories judged relevant to the current message — that keeps prompts small and answers on-topic.