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

# Monitoring Agent

> Use Monitoring Agent to investigate production alerts, attempt fixes, and automatically republish them. Ask Agent about your published app's logs and metrics.

Monitoring Agent helps you investigate problems with your published app. Enable
**Auto-fix** to let Agent respond to monitoring alerts and **Auto-republish** to
let it publish fixes automatically, or ask Agent about
production logs and metrics whenever you need help.

For charts, resource usage, and logs you can inspect yourself, see
[Monitoring](/features/publishing/monitoring-a-deployment).

## How it works

Monitoring alerts, **Auto-fix**, and **Auto-republish** are separate settings.
You can receive alerts without allowing Agent to change your code, or let Agent
attempt fixes while keeping control over publishing.

Automatic checks watch three signals: application error logs, HTTP errors, and
request latency. These checks don't cover every metric Agent can investigate.

With **Auto-fix** enabled, Monitoring Agent responds to alerts in the background:

**1. Investigate the alert**

Agent uses your app's production metrics, runtime logs, and code to
investigate the cause.

**2. Attempt and validate a fix**

When Agent identifies an actionable cause, it attempts a fix and validates
the change. An alert doesn't guarantee a fix. For example, an external
service problem may require action outside your app.

**3. Publish the fix**

With **Auto-republish** enabled, Agent can automatically publish its validated
fix. Otherwise, review the result in Agent chat and
[republish your app](/features/publishing/after-you-publish) to make the fix live.

> [!NOTE]
>
> **Auto-republish** is experimental and requires **Auto-fix**. Auto-fix alone
> doesn't authorize publishing changes to your live app. Some changes, including
> migrations and index operations, may be risky.

## Set up Monitoring Agent

Configure monitoring separately for each published app. Monitoring Agent runs
may incur [Agent usage charges](/billing/ai-billing#agent-billing). Track spending
on your [usage page](/billing/managing-spend).

**1. Open the publish settings**

Open the **Publishing** tool. For an existing published app, open its publish
settings. For a new app, configure monitoring during your first publish.

**2. Enable monitoring alerts**

For an existing published app, switch on **Monitoring alerts**. The change
saves immediately without republishing. Uptime checks are managed separately.

During your first publish, switch on **Enable Active monitoring** under
**Suggested add-ons**. This also enables uptime checks where available.

This example shows first-publish settings, with monitoring enabled and
**Auto-fix** and **Auto-republish** still off.

![First-publish settings with Enable Active monitoring on, Auto-fix off, and the experimental Auto-republish control visible and off](https://mintcdn.com/replit/DzLOj2ZVEmlWccT1/images/deployments/monitoring/monitoring-agent-setup.png?fit=max\&auto=format\&n=DzLOj2ZVEmlWccT1\&q=85\&s=86cb29ab23304c8314d7b7c9628039b7)

**3. Enable automatic fixes**

Switch on **Auto-fix** to let Monitoring Agent investigate
and fix monitoring errors.

For an existing published app, this change saves immediately. For a new app,
select **Publish** to apply your settings and start monitoring.

**4. Choose whether to republish automatically**

Enable **Auto-republish** to let Agent publish its fixes automatically.
Leave it off if you want to review and publish changes yourself.

For an existing published app, this change saves immediately. For a new app,
the setting takes effect from your first publish.

To stop automatic investigations and fixes, turn off **Auto-fix** in the same
settings. You can leave **Monitoring alerts** on to keep receiving alerts.
To keep automatic fixes but review changes before they go live, turn off
**Auto-republish** instead.

## Review alerts

Open the **Monitoring** tool and find the **Monitoring alerts** table:

* **Active**: Issues that currently affect your app.
* **Resolved**: Issues that stopped affecting your app within the last 30 days.

Expand an alert to see why it fired and how many monitoring passes detected it.
Alerts that Replit cannot currently evaluate show **Cannot measure**.

## Email preferences

Open **Settings** > **Notifications** and find **Active monitoring alerts** to
control your alert emails.

![Notifications settings with Production alert emails enabled and Active monitoring alerts set to Critical severity only](https://mintcdn.com/replit/DzLOj2ZVEmlWccT1/images/deployments/monitoring/monitoring-alert-notifications.png?fit=max\&auto=format\&n=DzLOj2ZVEmlWccT1\&q=85\&s=e0e4a3576a9b5dab2f802d5e452fc6a5)

Enable production alert emails if the severity selector is disabled. Choose
**Critical severity only**, **High severity and above**, **Medium severity and
above**, **All severity levels**, or **No emails**.

Account-wide recipient settings are separate from your personal email
preferences. In the account's advanced settings, open **Monitoring alert
recipients**. Authorized admins can choose **Off**, **Admins**, or **Custom**
recipients and a separate severity threshold.

Turning off emails doesn't turn off monitoring or **Auto-fix**.

## Ask Agent about production

You don't need to wait for an alert or enable **Auto-fix** to ask Agent about
your published app. Try prompts such as:

* "Why did my app return 500 errors in the last hour?"
* "Which endpoints were slowest today? Compare their p50 and p99 latency."
* "Check CPU and memory usage since my last publish."
* "Read the runtime logs for the revision that failed to start."

Include the approximate time, time zone, and affected endpoint when you know
them. Agent uses that context to narrow its investigation.

### Available telemetry

Agent can query more telemetry on demand than the three signals used for
automatic alerts. Available metrics depend on your deployment type and cloud
provider.

* **HTTP traffic**: Request count, request latency, 4xx and 5xx response counts, and availability.
* **Endpoint performance**: Request count, latency, and 4xx and 5xx response counts for each endpoint. Latency includes average, p50, p95, and p99 values.
* **Compute**: CPU and memory utilization, active instance count, startup latency, and instance uptime.
* **Storage and network**: Disk usage, disk utilization, and network egress.
* **Application errors**: Error log count and runtime logs filtered by severity or deployment revision.

## Next steps

* [Monitoring](/features/publishing/monitoring-a-deployment): Inspect your app's health, resource usage, and logs.
* [After you publish](/features/publishing/after-you-publish): Review and publish changes to your live app.
* [Manage spending](/billing/managing-spend): Track usage and control costs.