Guide · incidents

What happens between down and up.

When a confirmed state change fires, an incident opens. Everything that happens next — who acknowledged, what was tried, when it recovered — lands on one timeline you can read the morning after.

Lifecycle

OpenedAutomatically, on a confirmed DOWN (or a missed heartbeat). Alerts go out per the monitor's channels or alert plan.
AcknowledgedSomeone owns it — escalation steps stop advancing. One click in the app or POST /incidents/{id}/acknowledge.
EscalatedManually pushed to the next alert-plan step when you need more hands.
ResolvedAutomatically on confirmed recovery — recovery notices go to the same channels. Reopen if it was not actually over.

Severities

Incidents carry a severity from your own ladder — the defaults are conventional (SEV1 downwards), but the ladder is fully yours: rename, recolor, regroup via severities and severity groups.

The timeline: comments and attachments

Comments capture what was tried; attachments hold the screenshot of the weird dashboard or the log excerpt. Months later, the timeline is the postmortem's raw material — and the answer to “how did we fix this last time?”

Playbooks

A playbook is the restart procedure that otherwise lives in a wiki nobody can find at 4 a.m. — ordered steps, attached to the monitors they belong to. When an incident opens on such a monitor, the playbook is right there, and step completion is shared: everyone sees what has already been tried, in real time. Steps tick off per incident, so the playbook itself stays clean for next time.

Reporting

Incidents feed the reporting engine — incident analytics, downtime ledgers with verifiable content hashes, SLA reports with business-hours masking. White-label PDF reports deliver on schedule; see reports in the API reference.

Via API

The full lifecycle is drivable: open, acknowledge, comment, escalate, resolve, reopen — incidents reference. Every transition is also a signed webhook (incident.opened, incident.acknowledged, incident.resolved, …) — wire them into your own tooling instead of polling.