Skip to content
Uptime monitoring

know before your users dotell them before they ask

Checks that confirm a failure before they wake you, heartbeats for the jobs that run in the dark, and a public status page your users can subscribe to. So when something breaks, your inbox doesn't become the status page.

Marketing site

feedfa.st

99.98%

last 30 days · 193ms

90 days agoToday
Live data from the demo project’s status page, rendered by the same component your users would see.

confirmednot twitchy

An alert you learn to ignore is worse than none, so every default is built around not crying wolf.

  1. 1

    Confirmed in about twenty seconds

    A failed check is retried 5 and 15 seconds later before anything fires. A dropped packet dies on the retry, and a real outage reaches you inside the same cycle instead of an interval later.
  2. 2

    Never our outage, blamed on you

    When a check fails, we ask whether anything else on the internet answers. If nothing does, the fault is on our side: the check is recorded as unknown, your uptime is left alone and nobody is emailed.
  3. 3

    A 200 that says down is downPro

    Health checks often answer 200 while the database is gone. A monitor can read the answer: text that must or must not appear, or a JSON field that must equal, exist or stay under a number. If a rule fails, the check fails, and the alert says which rule and what came back.
  4. 4

    Alerts that carry the evidence

    The down email has the status line, safe response headers, the redirect chain and a slice of the body, so you can often diagnose it from the notification. Then a second email when it recovers, with how long it lasted. On Pro the same alerts land in Slack, Discord or Telegram. That evidence is yours alone: it never reaches the status page, the feed or the widget.

the quiet failuresthat never answer 500

Heartbeats, certificate warnings and private monitors come with every plan. A heartbeat counts as one of your monitors.

  1. Heartbeats for the jobs with no URL

    A nightly backup, a queue worker, an hourly import: when they stop, nothing returns a 500. Give the job a private check-in URL and add one curl to the end of it. If a check-in doesn't arrive within the period plus a grace you choose, the heartbeat goes down like any monitor: incident, email, chat alert, and recovery on the next check-in.
  2. SSL warnings weeks before the padlock breaks

    Every HTTPS monitor reads its certificate's expiry date from the handshake. You get one email at 14, 7 and 1 day before it runs out, so a renewal that quietly stopped working is a calm fix on a Tuesday, not a browser warning in front of your users. Nothing to set up.
  3. Private monitors stay off the status page

    Not every check is your users' business. Switch a monitor to private and it keeps checking and alerting you, but it's left out of the status page, the widget and the feeds, and its incidents are never emailed to subscribers. Heartbeats start out private.

a status pagethey can subscribe to

Public, on your slug, in your colours, with a ninety-day bar strip for every monitor.

  • Uptime over 24h, 7 and 30 daysComputed from real checks and cached, so the page stays fast when everyone is refreshing it.
  • Ninety days at a glanceOne bar per day, grouped by the visitor's local day, so a midnight outage lands on the day they lived it.
  • Incidents stay visibleResolved incidents remain, with how long they lasted. Hiding them is what costs trust.
  • Subscribers, double opt-inIncident and maintenance emails with one-click unsubscribe, from the same list your changelog emails go to. 30 on Free, 200 on Pro.
  • A status RSS feedFor people who would rather follow along without handing over an email address.
  • Times in their zoneThe page shows times in the visitor's zone, and every email writes the UTC offset out. Never a bare 02:00.

planned downtimeand the long kind

  1. Maintenance that isn't an outage

    Announce a window ahead of time and the affected monitors stop counting it: no incidents, no down alerts, the page shows Under maintenance in blue, and uptime % skips the window. Subscribers hear when it's scheduled, when it starts and when it's done.
  2. When it stays down

    An incident open for a day sends you one reminder. On Free, a monitor down for 3 days is usually a URL nobody fixed, so it's paused for you and you're told. Pro is never paused for you. Resume it whenever you like.

the same signalwherever people look

  • Checks every 5 minutes, or every minuteFree checks 1 monitor every 5 minutes. Pro checks 3 monitors per project, down to every minute.
  • Endpoints behind a tokenSent as a bearer header, a custom header or a query parameter, and stored encrypted. Never forwarded to a host your endpoint redirects to.
  • 30 or 90 days of historyEvery check is kept 30 days on Free and 90 on Pro, for the charts and the incident drawer.
  • The bar strip in your footerEmbed a monitor's strip on your own site with one attribute, and it stays live. Status and maintenance also come in the widget JSON, for showing them inline.
  • Slack, Discord and TelegramProDown, recovered, still down and certificate warnings, posted where your team already talks. Each channel picks what it gets.
  • Ask your agentProThe MCP server checks uptime right after a deploy and schedules the window before a migration.

Want the status page inside your own site instead? Start from one of eight status page examples and templates, each built around the same live widget.

before youpoint it at prod

Limits for each plan are on the pricing page.

Will a flapping endpoint flood my inbox?

No. A failure has to survive two retries, 5 and 15 seconds later, before an incident opens. The recovery email then waits ten quiet minutes, and if the endpoint drops again inside that window the same incident reopens instead of another down and up pair going out. You hear about real outages, once.

How am I told something is down?

By email on every plan, with the evidence attached: status line, safe headers, the redirect chain and a slice of the body, so you can often diagnose it from the notification. On Pro the same alerts also post to Slack, Discord or Telegram, and each channel picks which events it gets. Your status subscribers get a short, calm notice naming the monitor and the cause, never the diagnostics.

Can I monitor a cron job or a background worker?

Yes, with a heartbeat. Your job calls a private check-in URL when it finishes, for example by adding && curl -fsS to the end of a crontab line. If a check-in doesn't arrive within the period you set plus a grace of up to 6 hours, an incident opens and you're alerted. Periods run from 5 minutes to a week on every plan, and down to 1 minute on Pro.

Will you warn me before my SSL certificate expires?

Yes, on every plan and with nothing to set up. Each HTTPS monitor reads its certificate's expiry date every few hours and emails you 14, 7 and 1 day before it runs out, so a renewal that quietly broke gets fixed long before visitors see a browser warning.

Can I keep a monitor off the public status page?

Yes. Make it private and it keeps checking and alerting you, but it's left out of the status page, the widget and the status feed, and subscribers are never emailed about it. Ideal for admin tools, internal APIs and background jobs. Switch it back any time.

Can it check what the response says, not just the status code?

Yes, on Pro. Up to 10 rules per monitor, all of which must hold: text the body must or must not contain, or a JSON field that must equal, contain, exist or stay above or below a value. That catches the health check that answers 200 while the database is down. Rules read the first 512 KB of a GET or POST response, and if you ever move back to Free they're kept and pick up again when you return.

Can I monitor an endpoint that needs an API key?

Yes. A monitor can send a token as a bearer header, a header with a name you choose, or a query parameter, and the token is stored encrypted. A check that carries a credential never follows a redirect, so the token can't be handed to another host. For apps behind a sign-in form, point the monitor at a health endpoint: it's a stronger signal than a login page anyway.

How often can a monitor be checked?

As often as every minute on Pro, or every 5 minutes on Free, and as rarely as once an hour when that's all a page needs. Each monitor also takes a method (GET, HEAD or POST), the status code it should answer with, and a timeout from 1 to 30 seconds.

What should the monitor point at?

Ideally a health endpoint that does real work, touching the database or the cache, and answers 200 when healthy and 503 when not. Give each part of your system its own endpoint and its own monitor, so the status page names the part that broke instead of showing one red row.

Where do the checks run from?

From one location today, with a safeguard many checkers skip: before a failure counts, we confirm the wider internet answers. If it doesn't, the fault is on our side, so the check is recorded as unknown and never counts against your uptime or wakes you up. Checks from several regions are on the roadmap.

Point a monitor at your site in two minutes

Free for one project, with one monitor and a full status page. You'll know about the next outage before the first person emails you about it.