Monitoring · Available on Starter+

Planned downtime shouldn't page anyone or wreck your uptime numbers.

Schedule one-off or recurring maintenance across one or more monitors — checks keep running for accuracy, but alerts and uptime calculations exclude the window.

Start Free → See pricing

Free forever plan, no card required.

Why it matters

Maintenance windows

Without maintenance windows, every deploy that takes the API offline for two minutes either pages your on-call for nothing or quietly drags down your uptime percentage on a status page your customers can see. Neither outcome is fair to anyone. Paging on-call for a deploy you scheduled yourself trains the same alert fatigue as any other false positive, and a status page that dings your uptime number for planned work makes your reliability look worse than it actually is — a customer glancing at 99.7% instead of 99.98% has no way of knowing two minutes of that was a deliberate, announced deploy rather than something actually going wrong. Planned downtime and unplanned downtime are fundamentally different events and deserve to be treated that way, not lumped into the same alert and the same number. It's the difference between a deploy that's a non-event and one that quietly becomes a talking point in next week's status meeting. Nobody should have to choose between doing routine maintenance and keeping their uptime numbers honest.

One-off or recurring

Weekly or monthly rules, evaluated in your organization's own configured time zone. A one-off window covers a single planned deploy; a recurring rule covers something like a weekly Sunday-night maintenance cycle, set up once instead of having to be recreated by hand every single time it comes around. It's built to match how real maintenance actually happens, rather than assuming every window is a one-time affair. It's built around how maintenance actually gets scheduled in practice, not an idealized one-off case.

Real accuracy preserved

Checks keep running throughout the window — only alerts and uptime math are suppressed. That's the detail that makes maintenance windows genuinely useful rather than just a way to silence alerts: you still get a real recovery timestamp the moment the service comes back, which matters if the maintenance runs longer than planned. That's a deliberate design choice — silencing an alert should never mean silencing the truth about what actually happened. It's a small distinction that keeps the whole system honest without sacrificing anyone's sleep.

Incidents still tracked

Open/resolve history stays accurate for the record; just the outbound notification is dropped. If you ever need to look back and confirm exactly when a maintenance-window outage started and ended, that record is still there in full — it's only the page that would've gone to on-call that gets suppressed, not the underlying data. It's the kind of record you don't need until the one time you really do. It's the kind of detail that matters most exactly when you'd least expect to need it.

FAQ

Maintenance windows, answered.

Do checks still run during a maintenance window?

Yes — checks keep running as normal so you get an accurate recovery time. Only the outbound alert and the uptime percentage calculation are suppressed for the scheduled window, which means if the maintenance runs longer than planned, you'll still know exactly when things actually came back. That accuracy matters most on the rare occasion a routine deploy runs long. It's built to behave the same way you'd want any monitor to behave during planned work.

Can I schedule a recurring window, like every Sunday night?

Yes. Maintenance windows support one-off and recurring — weekly or monthly — rules, evaluated in your organization's own configured time zone, so a regular deploy window only needs to be set up once. After that, it applies automatically every week or month without anyone having to remember to recreate it. It's one setup, reused indefinitely, instead of a new entry every single week. It removes an entire category of recurring manual setup from the calendar.

What happens if I forget to end a maintenance window early?

Nothing dramatic — the window simply runs until its scheduled end time, and checks, alerts and uptime math all resume normally the moment it does. If the actual work finished early, the monitor is just quiet for a bit longer than strictly necessary; nothing is left in a broken or half-configured state. There's no cleanup step required on your end either way. The system doesn't need to be told the work is done — it just resumes on schedule.

Get Started

Maintenance windows. Free, starting now.

Start Free →