Memory that respects scope: per-user, per-bot, per-workspace

Few things make a customer feel less valued than having to repeat themselves. They told you their situation last week; this week they're explaining it all over again to the same company. Memory is what fixes that — an agent that recalls what it already learned. But memory is only an asset if it's scoped correctly. Remember the wrong thing in the wrong place, and you've created a problem instead of a convenience.

This post explains how Hania's long-term memory works, and why scope is the part that actually matters.

Why memory helps

Without memory, every conversation starts cold. The agent has no idea this caller has spoken to you three times before, or that they already told you their preference, or that their issue was half-resolved yesterday. With memory, the agent carries the useful context forward: it picks up where things left off, skips the questions it already knows the answers to, and treats a returning customer like a returning customer.

That continuity is a big part of what makes an agent feel competent rather than mechanical.

The scoping problem

Here's the trap: not everything an agent learns should apply everywhere. A detail about one customer should follow that customer — not surface in a stranger's conversation. A fact that's true for one bot's job might be irrelevant or confusing for another. And some knowledge really is shared across your whole operation.

If memory ignores these boundaries, two bad things happen. Useful context gets lost in the noise, and — worse — one customer's information can bleed into someone else's chat. Good memory isn't about remembering as much as possible; it's about remembering the right things in the right place.

Three scopes, on purpose

Hania scopes long-term memory along three lines:

• Per customer. What the agent learns about a specific person stays tied to that person, so it can recognize them next time without mixing them up with anyone else.
• Per bot. Memory tied to a particular agent stays with that agent's job, instead of leaking into unrelated bots.
• Across the workspace. Facts that genuinely apply to your whole operation can be shared, so every agent works from the same baseline where it makes sense.

Scoping it this way is what lets memory be both helpful and safe at the same time.

Memory you can see and control

Memory shouldn't be a black box. In Hania, the things an agent has remembered are visible in the dashboard, where you can review them, edit them, or delete them. If something was captured that shouldn't have been, you can remove it. If a detail is wrong, you can correct it. You stay in control of what your agents carry forward.

What it looks like in practice

A returning customer reaches your support agent. Because the agent remembers them, it greets them in context and doesn't re-ask for details they already gave. The specifics it recalls are theirs alone — another customer's chat is unaffected. And anything that's genuinely company-wide is consistent across every agent. The result is the thing customers actually want: they don't have to start over.

Getting started

Memory works alongside your agent's knowledge and tools. To see how the pieces fit together, explore chat agents or read about keeping personal data out of your logs.

Common questions