Detected in seconds. Posted with one click.
When a metric crosses its threshold, Observer drafts an incident automatically and emails it to your on-call. Publish or dismiss with one tap.
The old workflow: an engineer notices, types it up, posts it, and maybe remembers to send the subscriber email. Median time to publish runs 15 to 30 minutes, and by then customers have filed the tickets anyway. Observer detects the breach, drafts in seconds, and publishes on one click. Under a minute.
The drafting, done for you.
Everything between a breach and a ready-to-publish incident happens on its own. You arrive at the decision, not the paperwork.
- Automatic detection
- An incident is drafted the moment a metric's state change is confirmed. No one has to be watching the dashboard.
- Severity inferred
- The draft arrives pre-classified from the size of the breach, so a small dip and a hard outage do not read the same.
- One-click email
- On-call gets an email with Publish and Dismiss links. No app to open, no console to log into at 3am.
- Auto-expiring drafts
- If neither button is clicked within 24 hours, the draft expires on its own. Stale incidents never pile up.
- Maintenance windows
- Planned work is announced in advance and audit-logged, so expected downtime never trips an alarm.
- Per-customer routing
- Only the customers an incident actually affects are notified. Everyone else hears nothing.
Breach to published, step by step.
Six stages, only one of which needs a person. The rest is Observer doing the work a human used to do by hand.
- 01
Set the policy
Define an auto-incident policy on a metric: how severity maps to the breach, and how long it must dwell first.
- 02
The threshold breaks
The metric crosses its line and holds for the dwell window. A momentary blip never reaches this point.
- 03
Observer drafts
It infers the affected services, computes impact, and assembles a draft incident with a severity already set.
- 04
On-call is emailed
The draft goes to the designated address or Slack channel with Publish, Dismiss, and Edit links inline.
- 05
Publish or dismiss
Publish takes it live and notifies the affected subscribers. Dismiss discards it, and the decision is audit-logged.
- 06
Or it expires
If nobody acts within 24 hours, the draft is dropped automatically. No orphaned incidents linger.
Detection automatic, words reviewed.
Observer drafts; a person decides. Nothing customer-facing goes out without someone signing off on it.
- Whether the incident is published at all, or dismissed.
- The wording of anything that reaches a customer.
- Reclassifying the severity the draft proposed.
- Adjusting which services the incident is attached to.
What trips a draft.
A few policies in action, and the incident each one produces for review.
API latency over 1s for two minutes drafts a degraded-performance incident, ready to publish.
5xx error rate above 1% for 30 seconds drafts a major outage, severity already set high.
A service goes unreachable drafts a service-down incident with the affected customers already listed.