Status pages should read the metric.
Observer started from a simple frustration: the status page is the last place to know. It reports what an operator typed, minutes after the metric already moved. We thought the page should read the same signal the engineers do, and publish the verdict on its own.
One engineer, in the open.
Observer is built by Taha Al-Jody, an engineer who has run production systems and watched too many status pages lie by omission. The product reflects that experience: it trusts the metric your team already watches over a human remembering to flip a switch.
The data-plane is open source. The Observer Agent is Apache-2.0 and runs inside your own network. It evaluates every metric locally and pushes only the precomputed verdict outbound, so your raw telemetry, queries, and credentials never reach Observer Cloud. You can read every line on GitHub.
A page that follows the metric.
A status page has one job: answer whether the product is working right now. The honest answer lives in a metric, not in memory. Observer reads Prometheus, OpenTelemetry, CloudWatch, logs, databases, and network probes, applies the threshold you define, and flips the page within seconds. It recovers the same way, with no one typing anything.
Detection is automatic; communication stays human. Incidents auto-draft the moment a threshold is crossed, and a person approves before anything reaches customers. That is the line we hold: math decides what happened, people decide what to say.