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

# Enforcement layers

> How MergeGuide applies the same governance at each point in your workflow, so issues are caught early.

MergeGuide applies your policies at several points in your development workflow.
Each point is an **enforcement layer**. They all enforce the *same* governance —
the difference is *when* in your flow the check happens.

The point of multiple layers is **shift-left velocity**: catch an issue at the
earliest layer and it never has to be caught at a later one. A finding you fix in
your IDE never reaches a commit, a pull request, or an auditor.

## The layers

| Layer             | When it runs                | What you do                                                                                                                 |
| ----------------- | --------------------------- | --------------------------------------------------------------------------------------------------------------------------- |
| **IDE extension** | As you edit                 | Fix findings inline before you commit. See [Install the IDE extension](/install/ide-extension).                             |
| **MCP server**    | While an AI assistant works | The assistant checks a change against your policies as part of its loop. See [Install the MCP server](/install/mcp-server). |
| **Git hooks**     | On commit and push          | Staged changes are checked before they leave your machine. See [Install git hooks](/install/git-hooks).                     |
| **PR gate**       | On every pull request       | Each PR is evaluated and reported inline; blocking findings hold the merge. See [Set up the PR gate](/install/pr-gate).     |

## How they work together

You don't have to use every layer, and they don't duplicate work in a way that
slows you down — each is a chance to resolve an issue earlier:

1. The **IDE** gives you the fastest feedback, while you're still writing the code.
2. The **MCP server** keeps AI-written code in policy at the moment it's generated.
3. **Git hooks** are the last local checkpoint before code leaves your machine.
4. The **PR gate** is the shared, enforced checkpoint for the whole team.

Adopt the layers that fit your team. Many teams start with the IDE extension and
git hooks for fast local feedback, then add the PR gate to enforce policy for
everyone.

## Next steps

<CardGroup cols={2}>
  <Card title="Running scans" icon="play" href="/operation/running-scans">
    The scan commands behind these layers.
  </Card>

  <Card title="Reading findings" icon="list-check" href="/operation/reading-findings">
    Make sense of what a layer reports.
  </Card>
</CardGroup>