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

# Knowledge Base

> Understand how CodeRabbit Agent for Slack stores facts, separates public and private knowledge, and lets teams review fact history.

The Knowledge Base stores facts so CodeRabbit Agent for Slack can reference them in future conversations instead of starting from scratch. Knowledge follows Slack's privacy model: public channels write to shared workspace knowledge, while private channels and DMs keep their own isolated facts. Durable facts preserve knowledge, while [skills](/agent/skills) preserve reusable procedures. Together, they give the Agent both the context to remember and an established way to perform recurring work.

## Three ways to build knowledge

* **Explicit capture**: Tell CodeRabbit Agent to remember a fact from a thread or provide it directly. The most common way to add knowledge.
* **Knowledge proposals**: After a run, the Agent may offer to save durable facts it discovered. Public channel proposals can auto-apply; private channels require confirmation.
* **Background curation**: A weekly workflow refreshes existing knowledge from connected upstream systems automatically.

## Public vs. private knowledge

Knowledge boundaries follow Slack privacy boundaries.

| Slack surface                       | Knowledge behavior                        |
| ----------------------------------- | ----------------------------------------- |
| Public channels and shared surfaces | Use the global workspace Knowledge Base   |
| Private channels                    | Use a private conversation Knowledge Base |
| DMs and group DMs                   | Use a private conversation Knowledge Base |

<Info>
  Think of the global Knowledge Base as shared memory and private conversation knowledge as an isolated overlay with selective shared context available for reference.
</Info>

## Capturing knowledge

Ask the Agent to remember anything worth keeping: architectural decisions, operational facts, team conventions, or clarifications that would otherwise be lost. For example:

```
@coderabbit Remember that our payments service retries failed charges
exactly three times with exponential backoff. After the third failure
it marks the invoice as requires_manual_review.
```

CodeRabbit Agent saves the fact and references it in future conversations. You can review and manage saved facts in the web app.

<Info>
  In channels, you can also use the `/learn` slash command as a shortcut. Slash commands are not available in Direct Messages or threads.
</Info>

## What's next

<CardGroup cols={1}>
  <Card title="Durable skills" href="/agent/skills" icon="book-copy" horizontal>
    Learn how the Agent stores and reuses procedures alongside durable facts.
  </Card>

  <Card title="Working in Slack" href="/slack-agent/use-in-slack" icon="messages-square" horizontal>
    See how knowledge capture fits into normal Slack workflows.
  </Card>

  <Card title="Thread reviews" href="/slack-agent/thread-reviews" icon="file-search" horizontal>
    Inspect the run details and artifacts behind a conversation before deciding what knowledge should be preserved.
  </Card>
</CardGroup>
