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

# Test guardrails

> Check sample messages against an AI system's saved guardrail draft before you apply it.

In the guardrail editor, choose **Test guardrails**. The dialog checks a message against the saved draft. Nothing is sent to your AI system and no model answers: you see only what the guardrails would do.

**1. Save first**

The test uses the saved draft. If you have unsaved changes, the dialog says "Your unsaved changes are not included. Save the draft first."

**2. Write a message**

Enter up to 2000 characters in **Message to test**.

**3. Choose the direction**

In **Treat it as**, choose **What a person sends** or **What the model answers**. A guardrail that applies only to the other direction does not fire.

**4. Read the result**

Each message shows its outcome and which guardrails fired.

| Outcome | Meaning |
| - | - |
| **Nothing fired** | None of the guardrails fired on this text. |
| **Logged only** | A guardrail fired and logged it. The text would still go through. |
| **Blocked** | A guardrail fired and would stop this text. |
| **Masked** | A guardrail fired and would hide part of this text. The masked text is shown. |

The test log stays while you are on the page, even if you close the dialog, until you choose **Clear**. You can run up to 120 tests an hour.

To check the live guardrails against a whole set of attacks at once, use a [red-team run](/guardrails/red-team.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.