---
title: Memory
description: Long-term memory — the facts and preferences you want Ysra to carry across Knowledge conversations. Explicit, visible, scoped to an expert if you like, and deletable at any time.
---

**Memory** holds facts and preferences you want Ysra to carry from one Knowledge
conversation to the next. It is explicit: nothing is remembered unless you ask,
everything remembered is visible, and you can delete any of it.

![The Long-term memory page: a field with the example "Keep answers concise and call out assumptions explicitly", a kind selector set to Preference, an expert selector set to All experts, and a Remember button](../assets/screens/memory.webp){ width="1600" height="716" }

## Add a memory

**Workspace → Memory**, or ask in a conversation: *"Remember that our fiscal year
starts in April."*

| Field | Options |
|---|---|
| **Text** | The fact or preference, in your words. |
| **Kind** | *Preference* (how you like answers) or *Fact* (something true about your work). |
| **Expert** | *All experts*, or one expert only. |

## What gets remembered

- Only what you **explicitly** ask Ysra to remember, or clearly stable
  preferences you state.
- **Never** credentials, and **never** content retrieved from your sources —
  memory is not a copy of your company data.
- Ordinary conversation details stay in their conversation.

Memories are used only by experts allowed to read memory. Within one
conversation, Ysra always remembers the earlier turns anyway.

## Review and delete

The Memory page lists everything stored, by kind and expert. Delete any entry
and it's gone from future answers. The same list is on the phone under
**Settings → What Ysra remembers**.

!!! tip "Memory versus rules"
    Memory shapes **Knowledge** answers. Conventions for **Developer** sessions
    belong in [Rules & instructions](../developer/rules.md).