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

# Overview

> Catch bugs before they reach production and accelerate code‑review cycles.

cubic's AI reviewer delivers actionable feedback seconds after you open a pull request.

[Enable cubic on your repository](/ai-review/setup)

## Why modern teams choose cubic

* **Ship faster** – shorten review‑to‑merge time by up to 30 percent, letting your team focus on features instead of style nits.
* **Prevent production incidents** – surfaces regressions and edge-case failures before they reach customers, reducing outages and post-mortems.
* **Context-aware analysis** – examines your entire codebase and past merges to avoid noisy, one-size-fits-all warnings.
* **Actionable diff suggestions** – inline fixes you can apply with a single click.
* **Learns from your feedback** – replies or approvals adapt future reviews for the whole team.
* **Works out of the box** – zero configuration required; extend later with custom rules.

<Note variant="info">
  <span className="text-sm">
    cubic runs in an isolated container and never stores source code. See our <Link href="/account/privacy-security">AI security and privacy documentation</Link>.
  </span>
</Note>

## Key features

1. **Automated code reviews**

   * Scans pull requests for logic errors, style inconsistencies and potential vulnerabilities.
   * Adds structured comments and suggestions directly in the pull request.

2. **Codebase awareness**

   * Detects the libraries and frameworks in your repository, including version numbers.
   * References existing patterns or utilities when proposing improvements.

3. **Custom rules**

   * Enforce domain‑specific guidelines or internal best practices.
   * Violations are flagged automatically based on your configured rules.

4. **Adaptive feedback**

   * Learns from previous merges, review comments and dismissals.
   * Reduces false positives over time and tailors advice to your codebase.

## How it works

<Steps>
  <Step title="Enable AI review">
    Turn on AI review in your repository settings.

    <img src="https://mintcdn.com/mrge/eXeDFdBNkZEWK3p-/ai-review/images/ai-review-page.png?fit=max&auto=format&n=eXeDFdBNkZEWK3p-&q=85&s=3c0d84f27860d9cb5cf39c0b76dde119" alt="Toggle to enable AI review in repository settings" className="border border-zinc-800 rounded-lg" width="2874" height="1756" data-path="ai-review/images/ai-review-page.png" />
  </Step>

  <Step title="Open a pull request">
    cubic analyzes the changed files immediately after the pull request is opened.
  </Step>

  <Step title="Automated review">
    Inline comments highlight potential issues or recommended improvements.
  </Step>

  <Step title="Custom rules check">
    Any configured rules are evaluated and violations are noted.
  </Step>

  <Step title="Feedback and collaboration">
    Reply to an AI comment to refine future reviews.
  </Step>

  <Step title="Merge or iterate">
    Address the feedback, push additional commits or merge when ready.
  </Step>
</Steps>

## Manually trigger a new review

You can run the AI review again at any time:

* Select **Run AI review** from the pull‑request action menu.
* Or click the circular arrow icon on the right‑hand toolbar.
* Or tag **@cubic-dev-ai** in any PR and the AI will automatically run a new review and confirm it's processing.

<img src="https://mintcdn.com/mrge/eXeDFdBNkZEWK3p-/ai-review/images/manually-review-pr.png?fit=max&auto=format&n=eXeDFdBNkZEWK3p-&q=85&s=11726626f9ad9c7219aad08249616219" alt="Button to rerun AI review inside a pull request" className="border border-zinc-800 rounded-lg" width="2962" height="1724" data-path="ai-review/images/manually-review-pr.png" />

## Next steps

* **[Create custom rules](/ai-review/custom-rules)** – tailor the reviewer to your guidelines.
* **[AI review settings](/ai-review/ai-review-settings)** – configure library versions, ignore patterns and more.
* **[Providing feedback](/ai-review/feedback)** – fine‑tune the AI by responding to its comments.