Alerts
Alerts are available on Enterprise and Pro plans with Observability Plus
Alerts let you know when something's wrong with your Vercel projects, like a spike in failed function invocations or unusual usage patterns. You can get these alerts by email, through Slack, or set up a webhook so you can respond to issues.
By default, you'll be notified about:
- Usage anomaly: When your project's usage exceeds abnormal levels.
- Error anomaly: When your project's error rate of function invocations exceeds abnormal levels.
Vercel-defined minimum activity thresholds reduce low-volume noise and are not configured in alert rules. For details, see How error anomaly alerts trigger.
| Alert Type | Triggered when | Grouping |
|---|---|---|
| Error Anomaly | Fires when a route's five-minute error count is unusually high compared to its recent baseline and passes Vercel's minimum activity checks. By default, error anomalies track 5xx status codes, but alert rules can be configured for 4xx detections. | Route, HTTP status group |
| Usage Anomaly | Fires when your 5-minute usage is more than 4 standard deviations above your 24-hour average and crosses Vercel's minimum activity threshold. For Function invocations, Function duration, and Function CPU duration, usage combines metrics from Vercel Functions and Routing Middleware. | Metric |
Usage anomaly alerts support these metrics:
- CPU duration
- Duration
- Invocations
- Fast Data Transfer
- Edge requests
Use Configure alerts to create built-in alert rules, set notification destinations, and configure Slack or webhooks.
When you get an alert, Agent Investigation can run on its own to help you debug. Instead of digging through logs and metrics yourself, AI analyzes what's happening and displays highlights of the anomaly in your dashboard.
When you view an alert in the dashboard, you can click Enable Auto Run to trigger an investigation. This takes you to the Agents section in the sidebar, where you can set up investigations to run on new alerts. You can also click Rerun to start a new investigation.
Learn more in the Agent Investigation docs.
Vercel compares each completed five-minute interval with the route's trailing 24-hour baseline. These examples assume default sensitivity:
| 24-hour baseline | Current five-minute interval | Result |
|---|---|---|
| About 1,000 requests per interval, about 1% errors, and low variance | 1,000 requests and 100 errors | A high-impact anomaly triggers after the current interval |
| About 1,000 requests per interval, about 1% errors, and low variance | 1,000 requests and 40 errors | A lower-impact anomaly waits for the next interval, then triggers if the trailing two-interval validation passes |
| Similar average traffic and error rate, but with recent error spikes | 1,000 requests and 100 errors | The anomaly might not trigger if baseline variance keeps it below the threshold |
Actual results depend on baseline traffic, error rate, variance, and statistical confidence checks. Error counts alone do not guarantee an alert. The following interval does not need to contain another unusual spike, and alerts based on two intervals reflect data from both.
Was this helpful?
