Skip to content

Getting started with a monitor.

You add a target, pick how often it may be checked, and choose where a confirmed incident should go. The fastest interval depends on your plan.

How do you add a check?

  1. Create a workspace and pick a plan. Free allows five monitors and a check every five minutes.
  2. Add an HTTP monitor with the URL, the status codes that count as success, and a timeout.
  3. Add a TLS monitor on the same host if certificate expiry should warn you on its own.
  4. Add a heartbeat monitor for a job. Call the token URL from the job. You can rotate the token; the old hash stops working.
  5. Attach an email, Slack, Discord, or signed webhook so a confirmed incident has somewhere to go.

What does each state mean?

Monitor states
StateMeaning
pendingNo confirmed observation yet. A new or resumed monitor starts here.
upThe latest confirmed observation succeeded.
suspectOne failure. Not an incident yet.
downA second check confirmed the failure. The incident is open.
recoveringSuccesses have started. It is not up until enough of them arrive.
degradedPartial trouble: a certificate nearing expiry, or an HTTP check that succeeded but was slower than the threshold you set. A slow response does not open an incident.
maintenanceA maintenance window is active. Probes wait, and this time is left out of uptime.
pausedYou paused it. Checks stop until you resume, which returns the monitor to pending.
unknownNo fresh result. UPKira will not call this up.

When does an alert send?

When the incident opens, when it recovers, and when a TLS warning is due. A rule can mute any of those. Quiet hours hold the message until the window ends. The same incident is not sent twice to the same channel: each delivery has a dedupe key.

Read how targets and secrets are handled before you point a monitor at anything that is not a public URL.