Skip to main content
POST
Use this endpoint for operational faults: a job that gave up, a queue that stopped draining, a webhook that keeps failing. Repeated events for the same problem fold into one open alert and only the first one notifies. Use Create Incident instead when you want something customer-facing on a status page. An alert is internal until somebody promotes it.
Requires an API key with the alerts:write scope.

​
Body Parameters

​
string
default:"trigger"
trigger, acknowledge or resolve
​
string
What “the same problem” means to you, for example writeback:ezyvet:merchant-a576. Send the same key to fold repeats into one alert, and to acknowledge or resolve it later.Omit it and Kōdo groups on the summary with numbers and UUIDs stripped, so “job 41 failed” and “job 42 failed” land on one alert.
​
string
required
What went wrong. Required when event_action is trigger, max 200 characters.
​
string
default:"warning"
info, warning, error or critical. While an alert is open it keeps the worst severity seen.
​
string
Where the event came from, for example writeback-worker.
​
string
What is affected, for example ezyvet.
​
string
Service name or UUID to attach the alert to.
​
object
Free-form context. Shown on the alert and in the notification body.

​
Response

​
string
The alert this event landed on, or null if nothing matched.
​
string
triggered (new alert, team notified), deduped (folded into an open alert, nobody notified), acknowledged, resolved, or ignored (an acknowledge or resolve with nothing open).
​
string
The grouping key Kōdo used.
​
integer
How many events have folded into this alert.

​
Who gets told

A triggered outcome delivers to every notification channel with notify_on_alert enabled, and emails the team as a floor, capped at 10 alert emails per hour per organization. A resolved outcome goes to the channels only. notify_on_alert is on by default for every channel type except voice and SMS, which interrupt a person and cost money per delivery, so they opt in from the channel’s settings. A deduped outcome notifies nobody. That is the point: a fault that repeats every five minutes costs you one notification, not 288 a day.