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

# Prompts

> How to use the Prompt Management page in the dashboard.

The **Prompts** page is your central library for managing the prompts that power your AI features. Changes you make here take effect on live traffic within a few minutes, with no code deployments needed.

## Creating a Prompt

1. Click **New Prompt** in the top right.
2. Give it a descriptive **name** (e.g. `Customer Support Agent`).
3. Write your system prompt in the editor.
4. Optionally define a **model** and **temperature** to use by default.
5. Click **Save** — this creates the first version (`v1`).

## Version History

Every time you click **Save**, a new version is created. You can view the full version history by clicking the **Versions** tab on any prompt. To revert, click **Set as Latest** on any previous version.

## Using a Prompt in Code

Once a prompt is saved, it has a unique **Prompt ID** (a UUID). Reference it in your gateway requests using the `prompt_id` field — the gateway resolves it automatically at runtime.

```json theme={null}
{
  "model": "gpt-4o",
  "prompt_id": "your-prompt-uuid",
  "messages": [{"role": "user", "content": "Hello!"}]
}
```

See [Prompt Management](/docs/prompt-management) for the full integration guide.