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

# Memory

> Edit the workspace Memory the Cogno agent reads on wake — the persistent context behind better proposals.

**Memory** is a workspace-wide summary the Cogno agent reads on wake. It's a single
Markdown document that captures the durable context of your workspace, so the agent
works from an up-to-date picture instead of each event in isolation. Cogno keeps
Memory current on its own, and you can edit it directly whenever you want to correct
or add context.

## What Memory is

Memory holds the persistent context for one workspace — the background the Cogno
agent draws on when it makes sense of new events and proposes work. It's stored and
edited as **Markdown**, and it lives at **Settings → Memory**.

Two things keep Memory current:

* **Cogno updates it automatically.** As new events arrive from your connected tools
  (Slack, GitHub, meetings, and more), Cogno revises Memory to reflect what changed.
* **You can edit it directly.** Open the editor to fix, trim, or add context by hand.

<Note>
  Memory is workspace-wide: one shared document per workspace, not per person. See
  [Concepts](/docs/concepts) for how Memory fits the wider loop.
</Note>

<Tip>
  Memory describes *what is going on*. For rules Cogno must always follow — which
  channels to ignore, what a deliverable needs, assignment defaults — write
  [Instructions](/docs/instructions) at **Settings → Instructions** instead.
</Tip>

## Edit your Memory

<Steps>
  <Step title="Open Settings → Memory">
    Go to **Settings → Memory**. Cogno shows the current Memory rendered as Markdown,
    with a character count and the **Last updated** timestamp. If nothing has been
    written yet, you'll see **Memory is empty. Click Edit to add content.**
  </Step>

  <Step title="Select Edit">
    Select **Edit** to open the Markdown editor.
  </Step>

  <Step title="Write in Markdown">
    Edit the text directly. Memory supports Markdown — use headings, lists, and
    emphasis to keep it structured and easy for the agent to read. A workspace's
    Memory can hold up to **4000 characters**.
  </Step>

  <Step title="Preview or revert as you go">
    Select **Preview** to see the rendered Markdown, or **Revert** to discard your
    changes and restore the last saved version.
  </Step>

  <Step title="Save">
    Select **Save**. Cogno confirms with **Saved.** and updates the **Last updated**
    timestamp.
  </Step>
</Steps>

<Tip>
  Keep Memory concise and current. It's context the agent reads on every wake, so a
  tight, accurate summary beats a long one. Capture durable facts — active projects,
  key decisions, who owns what — rather than transient detail the agent already picks
  up from events. You have 4000 characters to work with.
</Tip>

## Reference

### Editor states

| State       | What you see                                                       |
| ----------- | ------------------------------------------------------------------ |
| **Loading** | Cogno is fetching Memory; the editor shows **Loading…**            |
| **Viewing** | Memory rendered as Markdown, with the **Edit** button.             |
| **Editing** | The Markdown textarea with **Revert**, **Save**, and **Preview**.  |
| **Saving**  | **Save** shows a spinner and is disabled until the save completes. |
| **Empty**   | No content yet — **Memory is empty. Click Edit to add content.**   |

### Limits

| Limit          | Value                           |
| -------------- | ------------------------------- |
| Format         | Markdown                        |
| Maximum length | 4000 characters                 |
| Scope          | One shared Memory per workspace |

## Related docs

* [Instructions](/docs/instructions) — standing rules Cogno always follows, next door at **Settings → Instructions**.
* [Concepts](/docs/concepts) — how Memory fits the events → agent → workers loop.
* [Getting started with Cogno](/docs/getting-started) — set up the loop end to end.