HTTP monitoring that checks more than a 200
Any URL, from every 30 seconds to once a day, asserted on status code, latency, body content and response headers — and confirmed from a second EU region before it wakes anyone.
An endpoint returning 200 is not the same as working
A load balancer serving a cached error page returns 200. An API answering in nine seconds returns 200. A deploy that replaced your JSON payload with an HTML login form returns 200. So a check that only reads the status line reports a healthy service while your customers cannot use it.
An okokumo HTTP check evaluates every assertion you give it, and counts as up only when all of them pass:
- Status code — exact, or a class such as any 2xx.
- Latency — a ceiling in milliseconds, so slow is a failure and not a footnote.
- Body content — a string that must be present, which is what catches the cached error page and the wrong deploy.
- Response headers — presence and value, for the cache, security and content-type headers that break quietly.
Interval and method
From 30 seconds to 24 hours. The floor depends on your plan — five minutes on Free, one minute on Starter, thirty seconds on Pro and Business — and the ceiling exists because some things are worth checking daily rather than constantly, like a report endpoint or a partner's callback URL.
Any HTTP method, with a timeout you set per check. IPv6 and self-signed certificates are handled rather than assumed away.
Two regions have to agree
Checks run from two independent EU probe regions. A failure only becomes an alert once a second probe confirms it, and a check only changes state after N consecutive failures — two by default, configurable. A single region losing its route to you does not page you at 3am.
The one deliberate exception: if a probe region is itself down, confirmation would mean never alerting at all, so the requirement relaxes rather than silently swallowing every failure.
How multi-region confirmation works →
What you see afterwards
Latency and uptime over day, week and month; every up and down transition with its timestamp; and a delivery log naming which alert went to which channel — including the ones deliberately not sent, with the reason. A monitor that quietly fails to page you is worse than no monitor, so that log is part of the product rather than a debug view.
Where the checks run from
France and a second EU region, on Scaleway. Your monitoring data does not leave the EU, and a DPA structured on Article 28 GDPR is available before you ask for it. Security & data →