> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trynito.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Memory

> Browser memory: facts recalled across chats, encrypted on your device.

Browser memory lets Nito recall facts about you across chats: preferences, project details, standing instructions. Memories are encrypted in your browser and scoped to your account and organization. They are never stored on the gateway.

## Turn It On

1. Open **Settings > Intelligence > Memory**.
2. Switch on **Workspace memory**.
3. Optionally add **Seed memory**: facts or standing instructions, one per line.

Memory needs a signed-in account.

## How It Works

* **Before a send**, matching memories are added as a private system message ahead of the request.
* **After every third completed reply**, Nito may extract a new memory from the recent turns. Encrypted Fusion skips extraction.
* **Per turn**, a composer toggle skips recall for that prompt.
* **Matching** uses an embedding model chosen by the gateway with no-store handling; the vectors and the memories live only in your browser.

**Require approval before saving memory** confirms each new memory before it is written.

## Privacy Levels

**Tier-aware memory privacy** is on by default. Each memory records the privacy level of the chat that produced it and replays only into chats at that level or higher.

* Confidential chats skip recall and extraction while the setting is on, because the embedding step is not itself Confidential.
* Encrypted chats use local retrieval in your browser, with no embedding request.

Switch the setting off for memory across every level. The same rule applies to Workbench Sources.

## The Memory Page

Open **Memory** from the main navigation.

* **Add a memory** by hand, in the third person ("the user prefers metric units"), with an importance from 1 to 10.
* **Import a document.** Text, Markdown, PDF, DOCX, PPTX, and XLSX are parsed in your browser and chunked into memories. Up to 10 MiB per file, 50 documents per organization, 500 chunks per document.
* **Saved memories** lists every record with its importance and source. Deletion is permanent.
* **Imported documents** can be paused and resumed as a whole.
* **Clear all** removes every memory in the organization.

## With Your Other Data

* **Purge chat data** in Settings does not touch memory; use **Clear all** on the Memory page.
* Memories are included in [encrypted backups and live sync](/web-app/backups-and-sync).
* Incognito chats do not write memories.

## Next

<CardGroup cols={2}>
  <Card title="Workbench" icon="layer-group" href="/web-app/workbench">
    Context you attach on purpose.
  </Card>

  <Card title="Privacy in the App" icon="user-secret" href="/web-app/privacy-in-the-app">
    Levels and Incognito chats.
  </Card>

  <Card title="Backups and Sync" icon="cloud-arrow-up" href="/web-app/backups-and-sync">
    Memories on another device.
  </Card>

  <Card title="Data Retention and Visibility" icon="eye" href="/privacy/data-retention-and-visibility">
    What stays on your machine.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.