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

# Scan with AI-powered detection (beta)

This page provides step-by-step instructions on enabling and running an AI-powered scan. For details on what AI-powered detection can uncover, known limitations, and beta considerations, see [AI-powered detection overview](/semgrep-code/ai-powered-detection-concepts).

## Prerequisites

To run Semgrep Code's [AI-powered detection](/semgrep-code/overview#ai-powered-detection-beta-feature), ensure that you meet the following requirements:

* You have added your projects to [Semgrep Managed Scans](/getting-started/quickstart-managed-scans#add-projects-to-semgrep-managed-scans). Look for the `managed-scan` tag in the [**Projects** section of the Semgrep AppSec Platform](https://semgrep.dev/orgs/-/projects/scanning).
* You have enabled [Semgrep Multimodal](/semgrep-multimodal/getting-started#enable-multimodal) for your organization.

## Enable or disable AI-powered detection

This feature is enabled by default for all Semgrep Multimodal users.

To enable or disable AI-powered detection in Semgrep AppSec Platform, go to [**Settings** > **Code**](https://semgrep.dev/orgs/-/settings/general/code) and then toggle **AI-powered scanning** on or off.

## Scan with AI-powered detection

<Steps>
  <Step>
    Log in to Semgrep AppSec Platform.
  </Step>

  <Step>
    In the **navigation bar**, click on **Projects**.
  </Step>
</Steps>

To scan the default or main branch:

<Steps>
  <Step>
    Choose the projects by selecting the checkboxes next to their names. This enables the **Run a new scan** drop-down menu.
  </Step>

  <Step>
    Click **Run a new scan > AI-powered detection**.
  </Step>

  <Step>
    A dialog appears that displays the number of projects that were selected for scanning. Click **Scan** to begin.

    * If you would like Semgrep to automatically perform an AI scan on these projects every week, select **Enable weekly scans**.
  </Step>
</Steps>

To scan a non-default branch:

<Steps>
  <Step>
    Click **Details** for your project of interest. On the project's **Details** page, click **Run a new scan** and choose **AI-powered detection.**
  </Step>

  <Step>
    In the dialog, enter the name of the branch you want to scan.
  </Step>
</Steps>

## View findings

Findings generated by AI-powered detection scans are part of [Semgrep Code findings](/semgrep-code/findings) and are listed on the **[Code](https://semgrep.dev/orgs/-/findings)** page. You can use the filters icon to filter for **AI-powered scan findings**.

The findings card indicates whether a finding was detected by an AI-powered scan or a Rule-based scan.

## Add additional context to AI-powered detection scans

<Note>
  **INFO**

  Only **Admins** can upload context documents to Semgrep projects.
</Note>

By uploading project-specific context such as design documents, threat models, or instructional markdown, you can provide additional information for Semgrep to use during AI-powered scans. This enables Semgrep to show higher-impact findings and reduce false positives based on how your application is designed and used.

To upload a project-specific context document:

<Steps>
  <Step>
    Log in to Semgrep AppSec Platform.
  </Step>

  <Step>
    In the **navigation bar**, go to **Rules & Policies > Memories**.
  </Step>

  <Step>
    Go to the **Documents** tab and click **Add document**.
  </Step>

  <Step>
    Drag the document to the **File upload** box or click **Choose a file** to select and upload your context document.

    Optionally: Add a **Description** of the document. This information will be used as additional context for AI-powered detection scans.
  </Step>
</Steps>

The finding **Details** page references the uploaded context document under the finding description.

For an in-depth understanding of how AI-powered detection works, see [AI-powered detection: concepts, limitations, and FAQs](/semgrep-code/ai-powered-detection-concepts).
