# Alerts, incidents & API drift - ApiTraffic

> ApiTraffic compares every endpoint with its own normal, opens one incident per real problem and notifies Slack, email, PagerDuty, Jira or a webhook. Plus API drift detection, release markers and post-deploy regression tracking.

Source: https://dev.apitraffic.io/monitoring

Alerts, drift & releases

# Find out it broke before anyone files a ticket.

ApiTraffic learns what normal looks like for every endpoint, then opens an incident when one gets genuinely worse, flags the deploy that changed your API’s shape, and shows which queries got slower after each release. Alerts are included on every plan.

[Start your free trial](https://dev.app.apitraffic.io/signup?src=monitoring-hero)[See pricing](/pricing)

Alerts & incidents

## One incident per problem. Zero 3am false alarms.

Every five minutes, ApiTraffic checks the last hour of every endpoint against its own normal. A change has to be a multiple of normal, big in real terms and statistically real - all three - before anyone is told. Related alerts are grouped into a single incident, so a bad deploy is one page, not forty.

- **Channels:** Slack, email, webhooks, PagerDuty and Jira
- **Routing** with escalation steps, quiet hours, digests and silences for maintenance

- **A timeline for every incident:** each alert, each notification and why it was sent
- **Acknowledge, snooze or resolve** from the dashboard, and channels hear about it

What counts as broken

## Built to page you only when it matters

Thresholds tuned so that a busy endpoint with a few errors a day stays quiet, and a real regression is reported in about ten minutes - or straight away when it is severe.

### Server errors

Per endpoint at 3× its normal rate and at least 2 points higher, and across the whole bucket at 2× and 1 point.

### Client errors that are your fault

4xx spikes count only when they are spread across consumers, so one client sending bad requests never pages you.

### Latency

p95 at 1.5× normal and at least 150 ms slower. Critical when it triples and adds a full second.

### Breaking API changes

A field removed or retyped in a response opens an incident too, straight from drift detection.

### Normal means your normal

The higher of the last hour and the same hour last week, so the Monday 9am spike is not an incident every Monday.

### Never silently broken

If every channel a route sends to is failing, ApiTraffic opens a critical incident about that and emails your owners.

API drift detection

## Your API changed shape. Know before your consumers do.

Drift detection learns each endpoint’s normal request and response shape, status codes, latency and dependencies from its own traffic. Every hour it compares recent traffic with that baseline and raises a finding - with the diff, the baseline and the requests that prove it.

- **Removed and retyped fields** caught from the bodies you already log
- **Correlated with deploys,** so each finding names the release that likely caused it

- **Diagnose with AI:** a grounded prompt your coding agent can act on
- **Observe-only:** drift never blocks a request

### Breaking

A field your consumers read was removed or changed type, in a request or a response.

### Additive

New fields appeared - usually fine, worth knowing before someone depends on them.

### Behavioral

The status code mix, error rate or latency moved away from the endpoint’s baseline.

### Dependency

An upstream the endpoint calls started failing or slowing down.

The Drift page for the sample API: open findings by endpoint, with change type, severity and the samples behind each one.

Release markers

## Every deploy, on every chart

Mark deploys, incidents, rollbacks, scaling and config changes, and ApiTraffic draws them on your traffic and latency charts. When errors jump, the change right before it is impossible to miss.

- From your CI pipeline with one API call - GitHub Actions, GitLab CI, Jenkins and more
- Automatically from linked GitHub, GitLab and Bitbucket repositories

- From the app, or by asking your AI assistant over MCP
- Add the commit, and Fix with AI names the deploy most likely to blame

```
# After a deploy: one call, and the marker lands on every chart
curl -X POST "https://api.apitraffic.io/v1/accounts/$APITRAFFIC_ACCOUNT_SID/events" \
  -H "Authorization: Bearer $APITRAFFIC_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "deployment",
    "name": "v2026.10.24",
    "metadata": { "commitSha": "'"$GIT_SHA"'", "repoFullName": "your-org/store-api" }
  }'
```

The sample API's last hour: the 17:40 deploy and the 17:45 incident drawn on throughput and response time.

Regression tracking

## Did that release make anything slower?

Pick a deploy and ApiTraffic ranks the operations and endpoints that got slower after it - p95 before, p95 after, the change and the calls - weighted by traffic, so a small slowdown on a hot path outranks a big one nobody calls. Open any row for its latency trend, heatmap and a before-and-after verdict.

- **Compare around any deploy** or any moment you pick
- **Slowest real examples** one click away, each with its waterfall

- **Ask your assistant:** the MCP tool `regressions_since_deploy` checks a release for you

## Get told about the next regression before your customers are.

Install the SDK, ship as usual, and let every endpoint learn its normal. Alerts, drift and release tracking start working with your first traffic.

[Start your free trial](https://dev.app.apitraffic.io/signup?src=monitoring-closing)

Free trial · No credit card · Set up in minutes