Uptime Kuma hosting
Hosting here means the Uptime Kuma you already know, running on our fleet under your own account: your own instance with your own database, a confirmation vote across three regions standing in front of every down alert, and status pages mirrored onto separate infrastructure. The software stays open, the exit stays open, and the price stays flat.
What hosted means here
A hosted Kuma can mean many things across the market, so this is the exact shape of ours, in the open, before any talk of price.
Every customer runs in a separate Uptime Kuma instance with a separate database, so your monitors, your history, and your alert settings live in a place that belongs to your account alone and never mixes with another customer's data.
When your instance reports a monitor down, probes in New York, Frankfurt, and Singapore check the same address at the same moment, and the alert goes out only when at least two of the three probes also fail to reach it. When the vote fails to confirm, the alert is held back and the suppression is written into your own instance as a log entry you can read.
Status pages are copied off the tenant host onto separate infrastructure on a schedule, so the page your customers read during an incident never sits on the machine that is having the incident. Our own public status page runs on exactly this mirroring.
The code that runs this service is published at github.com/heavyreading/downforreal under the MIT licence, with the upstream notices intact. You can read exactly what watches your sites, including the confirmation layer, before you trust it with an alert you sleep next to.
Your data lives in a SQLite database inside your own instance. If you ever decide to self-host again, you can take the same public software and your own data and run the stack on your own server, which means choosing hosting here never has to become a lock-in you discover later.
Self-host or hosted
Self-hosting Uptime Kuma is a genuinely good option for plenty of people, and any page that pretends otherwise is selling you something. Here is where each side actually stands.
| Question | Self-hosted Kuma | DownForReal hosting |
|---|---|---|
| Price | Self-hosting wins. The software is free and open source, and you pay only for whatever server you run it on. | One flat price: $10 a month, or $75 a year, with the fleet and its upkeep inside it. |
| Control | Self-hosting wins. Your server, your versions, your maintenance windows, your rules, end to end. | The fleet is ours to run. Your instance, your monitors, your data, and your settings are yours. |
| Your time | Updates, TLS certificates, backups, and the health of the monitoring box itself all land on you, including the awkward question of what watches the watcher. | Handled centrally on our side: the host, the updates, the certificates, and daily backups, with fleet improvements rolling out without an evening of your time. |
| Multi-region confirmation | Upstream closed the request as not planned. Issue #1915 asked for multi-location checks in 2021, and the standing workaround is a second install on another network, synced by hand. | Built into the alerting path: probes in New York, Frankfurt, and Singapore vote on every down result, and at least 2 of the 3 must agree before the alert fires. |
| Status page during an outage | The page lives wherever your instance lives, which is usually the same box whose failure the page exists to announce, unless you arrange a second home for it yourself. | Status pages are mirrored onto separate infrastructure on a schedule, so the page stays readable while the host behind it is down. |
| Leaving later | You already hold the software and the data, because both were always on your machine. | The fork is public under MIT and your data sits in a SQLite database inside your own instance, so leaving means taking the same stack with you. |
The short version runs like this: if you enjoy running a small server and your time costs you nothing, self-host, because you will pay less and control more. Hosting earns its price when your time is the scarce thing, or when the confirmation vote is the feature you actually came for, because that vote is the one piece upstream has declined to build. The free checker runs the same vote in public, so you can watch it work before deciding anything.
Pricing
There are no per-monitor meters and no per-seat charges to model before you can predict the bill, and the trial runs the full plan for 14 days.
Or $75 a year on the annual plan, which brings the same plan down to $6.25 a month across the year.
Your own instance and database, up to 100 monitors at a 60-second minimum interval, down alerts confirmed by the three-region vote, status pages mirrored off-site, the UptimeRobot importer for an existing list, and email support from the operator who runs the fleet.
Coming from UptimeRobot? The migration guide walks through the import step by step, including the one manual step.
Questions
No official status is claimed, and the product carries its own name, DownForReal, for exactly that reason. The service is powered by Uptime Kuma, the fork it runs is public on GitHub under the MIT licence, and the attribution on every page says so in plain words.
Your own Uptime Kuma instance with your own database, up to 100 monitors at a 60-second minimum interval, the three-region confirmation vote standing in front of your down alerts, status pages mirrored off the host that runs your monitors, daily backups, the UptimeRobot importer if you are bringing a list with you, and email support from the operator who runs the fleet.
Your instance checks from its home on the fleet, the way stock Kuma does. When it reports a monitor down, probes in New York, Frankfurt, and Singapore check the same address at the same moment, and the alert goes out only when at least two of the three probes also fail. When the vote fails to confirm, the alert is held back, and the suppression is logged inside your own instance where you can read it later. The free checker runs the identical vote against any public address, with no account needed.
Plenty of people should, and the comparison section above gives self-hosting its wins on price and control without qualification. The two things hosting adds are time and the vote: somebody else keeps the monitoring box itself alive, updated, and backed up, and every down alert gets confirmed from three regions, a capability upstream closed as not planned in issue #1915. If neither of those matters to you, your own server is the better buy.
Yes, and the route is built out of the open pieces rather than a promise. The fork is public under the MIT licence, and your data lives in a SQLite database inside your own instance, so the same software and your own history can be running on your own server whenever you decide the hosting has stopped earning its price.
The full plan for 14 days, with your own instance, the confirmation vote, and mirrored status pages all live, and a 300-second minimum check interval during the trial period against the 60-second floor on the paid plan. Bring an UptimeRobot list with you and the importer runs during the trial as well, so the two weeks can run side by side against your old setup.
The free checker shows the three probes voting on any public address in seconds, and the trial then runs the full hosted instance for two weeks. Between those two, you can judge the hosting on evidence gathered from your own services rather than on this page.