A single flaky route does not page anyone.

Perstat confirms an incident across probes on six continents before it pages you. HTTP, TCP, and DNS checks down to 10-second intervals, with on-call, status pages, and SLA reports in one product. No APM, no log management, no tracing, and we say so up front.

The mechanism, not the benefit

You have been paged at 3 a.m. by a monitor that watched one route hiccup from one location. That is not a signal. That is a coin flip with your sleep.

Perstat does not alert on a single failed check. A monitor runs from probe regions on six continents: North America (na-ewr), South America (sa-gru), Europe (eu-fra), Africa (af-jnb), Asia (as-sgp), and Australia (oce-syd). An incident opens only when more than one region confirms the failure, and within each region 75% of probe nodes have to agree before the region votes at all. One flaky route, one saturated transit link, one bad resolver on one continent does not reach your phone.

And nothing below quorum is silently swallowed. A forming incident is visible in the product with the first failing region named, and sub-quorum divergences (one region drifting, one IP family failing, one node disagreeing) land on a watchlist you can investigate on your own schedule instead of at 3 a.m.

The point is that this is a rule, not a model. There is nothing to tune and nothing to second-guess. The incident’s start time is the moment quorum confirmed, not the first failed check and not a rounded minute. The same rule that keeps you asleep also floors your SLA later: the window in the report runs from confirmation to confirmed recovery, the same boundary you were paged on. One record, no drift between the story your customer sees and the story your auditor gets.

What Perstat is not

No APM. No log management. No tracing. If a request is slow inside your service, Perstat will not tell you which span ate the time, and it does not pretend to.

What it covers is the layer from the URL down to the host:

• 12 check types, from HTTP with assertions through DNS hygiene (SPF, DMARC, CAA), domain expiry, TLS certificates, SMTP and IMAP, down to ping and traceroute. The check-type reference explains each.
• Host agents for Linux and macOS: CPU, memory, disk, and services, self-updating so you are not chasing agent versions.

That is the whole scope, on purpose. Solve availability and its evidence completely, and leave deep tracing to the tools built for it. The full tour is on the product page, no signup.

Pains, mapped to what handles them

• Tool sprawl. Checks in one tool, a status page in another, on-call in a third, none of them sharing data. Perstat is four pillars in one product: monitoring, status pages, incident response and on-call, and evidence. See all four.
• Alert fatigue. The quorum above. An alert means more than one continent agreed before your phone lit up.
• SLA math by hand. Curated exclusion windows with real boundaries, announced maintenance marked as such, availability floored and never rounded up. Hand the sample SLA report to a customer and let them recompute it.

What lands on the record, and what does not

Announced maintenance windows suppress alerts for their duration and still stay on the record; planned work is visible, not counted against you. Measured downtime is kept separate from the incident story: a false positive can be discarded with a reason, with the uptime effect previewed first. The audit half of that story, corrections logged and reversible, retention windows from 7 days to 2 years, lives on the leadership page.

Your data leaves the way it came in

Before you spend a week on setup, the exit matters more than the onboarding. Your records are yours. Availability evidence is exportable on request today, self-service export is the stated next step, and if you leave we delete on request rather than hold your history hostage.

The control plane runs in the EU, operated by a German company with no US parent, so there is no US CLOUD Act reach into it. The probe nodes on six continents are measurement points; the data and the controls stay in the EU. The specifics, including retention and an honest account of what we do not have yet, are on the security page.

Switch once

If you run on-call in Opsgenie, the calendar already made this decision for you. End of sale was June 2025, and the service shuts down on April 5, 2027, with customer data deleted. The migration is due regardless of who you pick.

Be honest with yourself about the effort. The first check is running in minutes. Rebuilding on-call rotations, escalation policies, and re-pointing every alert source is days of real work, and you should run both systems in parallel for a full rotation before you cut over. Anyone selling you “effortless” has not moved a live on-call setup. The concept-by-concept mapping is on the Opsgenie comparison, and the migration guide has the order that keeps you covered. Coming from somewhere else? Better Stack, UptimeRobot, Statuspage.

Questions a peer will check in five minutes

All-in-one usually means mediocre at everything. Fair prior. The depth is real per pillar: escalation from push to SMS to phone call with acknowledgment; status pages with custom domains, double-opt-in subscribers, and white-label from Command; SLA reports with real window boundaries. And the scope sentence still holds: no APM, no log management, no tracing.

Are 10-second intervals and six continents real, or marketing rounding? Intervals are configurable down to 10 seconds, and the floor depends on the plan: Free 300s, Pulse 60s, Sentinel 30s, Command 15s, Enterprise 10s. The regions are named above, not a world map with dots. We run Perstat on Perstat, and our own status page goes public with the launch; from that day, check the claim there.

Is the free tier honestly usable? Ten monitors at 300-second intervals plus a public status page, at 0 EUR with no time limit. That is a real free tier for a side project, not a 14-day clock. The full ladder, net and excl. VAT: Free 0, Pulse 29, Sentinel 89, Command 249, Enterprise from 690 EUR per month. Every number is on the pricing page, no sales call.

We are on Android. Alerting is platform-neutral: push, SMS, and a phone call reach every phone. The native iOS app and Apple Watch, currently in TestFlight, are a bonus for the Apple half of your team, not a requirement to run on-call.

Is there an API? Can I manage monitors as code? Not yet, and we would rather say so before you sign up. Today you configure monitors in the web app and the iOS app. A public, documented API with scoped keys is on the roadmap, not shipped. It shows up on the changelog when it lands, dated like everything else.

Start with one check

Point one monitor at one URL and watch what it records. That is the fastest way to judge the rest.

Start monitoring free. Or see pricing first. Or take the tour, no signup.