Down means down
A site is marked down only when probes on three continents and three different networks agree, twice in a row. If it’s down, it’s down. If it isn’t, you sleep.
650,000 free credits every month. No card, no seats, no subscription.
Monitors
Checked from 15 probes on 6 continents.
One failure is a rumour. Agreement across continents is an outage.
Most monitors alert the moment one location times out. That location may have a routing problem, a full disk or a noisy neighbour. We treat a failure as a question: every healthy probe checks it again, from every continent, before anyone is paged.
Scheduled check
On every interval six probes on at least four continents check your site at once. That whole check costs one credit.
The whole fleet checks again
The moment one probe sees a failure, every healthy probe checks your site again, about every 20 seconds. Confirmation rounds are free and never touch your balance.
Quorum, then alert
Down means two thirds of healthy probes agree, from at least three continents and three networks, and the failure lasts. Then you’re paged, with what broke: refused connection, expired certificate, wrong DNS answer, or a route problem in one region.
Built like the thing it watches.
A monitor you can't trust is worse than none. So the probes watch each other, the central watches itself, and nothing about the design depends on one machine.
- probe fleet
- 15 dedicated servers in 15 cities on 6 continents, at 5 different providers. Static IPv4 (IPv6 where the provider offers it), published reverse DNS.
- quorum
- Two thirds of healthy probes, from at least 3 continents and 3 networks, two rounds in a row. A failure seen from one region only is shown as a partial outage, never as down.
- probe self-check
- Each probe tests known-good canaries. A probe that can’t reach them is quarantined and its votes don’t count. Below 8 healthy probes we never page; we tell ops instead.
- central outage
- Probes cache their job list and keep checking. Results are buffered in memory and sent when central returns.
- dead-man's switch
- Central's scheduler pings an external heartbeat every tick. If it stops, we're paged by someone else's monitor.
- data
- PostgreSQL, backed up every night, encrypted, to off-site storage. A full restore is tested automatically every week.
- retention
- 14 d raw per-probe results · 1 y hourly rollups (uptime, p50, p95) · incidents forever.
- safety
- Every target resolved once, refused if any address is private, connected to by IP. Redirects re-validated. Body capped at 256 KB.
- abuse gate
- Any public host gets HTTP/TLS on 80/443 and ping, every minute. Verify a domain via DNS TXT or a file to unlock 15 s, any port, TCP and SMTP. Owners can opt out at /probes.
- jurisdiction
- Operated from Greece. Your data stays in the EU; only the checks run worldwide. GDPR export and deletion built in.
The checks that matter, done properly.
Confirmed by the whole fleet
One failure starts free re-checks from every healthy probe. Nobody is paged for one city’s bad route, and the alert lists what each probe saw.
HTTP(S) and keywords
Status codes, keyword present or absent, redirects followed and re-checked. Catches a WordPress database-connection error page served with a 200. For APIs on your verified domains: any method, custom headers, a request body and Basic auth.
TLS and expiry
Certificate days read from the same connection as the check, on 443 or any port, including SMTP STARTTLS.
Ping, TCP, SMTP, DNS
Ping any public server, IPv4 or IPv6. Ports, mail banners and DNS answers for domains you've verified.
Cron jobs and backups
Heartbeat monitors: your job calls its own address when it finishes, or reports a failure with the reason. Silence past the grace period pages you.
Domain expiry and slow responses
Every monitored domain is checked for expiry daily, at no cost. Optional slow-response alerts use the median of all probes, so one slow city can't trigger them.
Blocked, not down cloudflare
A WAF challenge or a Cloudflare block page is reported as “blocked”, never as down. Connect your Cloudflare account and we keep the allow-list rule in place for you, or follow the guide and add it yourself.
Partial outages and geo-blocking
Unreachable from Asia but fine everywhere else? You see exactly that, with the region. If your site blocks a continent on purpose, we notice and let you take those probes out of the vote.
Dev / stage mode
Staging sites are watched without paging anyone. You see the history, nobody wakes up.
Import from where you are
Paste a read-only key from UptimeRobot or updown.io, or upload an Uptime Kuma backup file or a CSV. You see every monitor we found and the monthly cost before anything is created. Export the same CSV whenever you like; nothing is locked in. How switching works.
Groups and bulk changes
Put one customer’s monitors in one group: its own section on the dashboard, its own status page, its own alert channel, its own maintenance window. Select every monitor that matches a filter and change the interval, channels or group in one go; the credit cost shows before you confirm.
Maintenance windows
Once, daily, weekly or monthly, for one site, a whole group or everything. Checks keep running; nobody is paged, and it doesn’t count against your uptime.
Evidence in every alert
What each probe saw: resolved IP, status, timing, error. Forward it to your host and skip the argument.
Public status pages
Pick the monitors or a whole group, name them, publish. Visitors see confirmed facts only: operational, partial outage, outage or maintenance, with 90 days of history. Your logo and colours.
Email, Slack, Teams, Discord, Telegram
Alerts go where you already look, each on its own terms: Telegram at once, email only after 10 minutes, reminders every 30 while it’s down. Also Google Chat, PagerDuty, phone push through ntfy and Pushover, and signed webhooks.
For hosting providers enhance
Connect your Enhance panel under Integrations and every website is monitored, paused when suspended, removed when deleted. Synced every 10 minutes.
Pay for checks. Nothing else.
One credit is one scheduled check, whatever runs behind it. Confirmation rounds are free. Every account gets 650,000 credits a month for free. Above that, buy a pack when you need one; there is no subscription.
| Pack | Price | Per million |
|---|
- 650,000 free credits every month, for every account. No card.
- Free credits are used first and reset on the 1st. They don’t roll over.
- A 1-minute monitor uses 43,200 credits a month.
- Failed checks cost the same as passing ones. Confirmations cost nothing.
- An email when about a week of credits is left. At zero you get 7 days’ grace, then monitors pause, with a message on every alert channel. Nothing stops silently.
- Bought credits are valid for 24 months. Prices are before VAT; VAT is added at checkout where it applies.
What 50 monitors at 1 minute cost.
Public list prices, checked . Monthly cost on each vendor's cheapest plan that covers it.
Full comparison with features, locations, confirmation policy and every source.
Plain answers.
Where are the probes?
Frankfurt, London, Stockholm, Amsterdam, Warsaw, New York, New Jersey, Dallas, San Francisco, São Paulo, Singapore, Tokyo, Fujairah, Sydney and Johannesburg. Names and IPs are on the probe page.
Will you ever miss a real outage?
Confirmation adds about a minute before we page. In exchange, every page is real. If one region can’t reach you, you see a partial outage with the region, and can choose to be alerted for that too.
How many minutes after a site goes down do I hear?
At most the check interval to notice it, plus about a minute to confirm it. Checked every 5 minutes: within about 6 minutes. Every minute: within about 2. Every 15 seconds: in a little over a minute. The alert itself leaves within seconds of confirmation.
My site is behind Cloudflare. Will it work?
Yes. We publish our probe IPs and show the exact WAF rule that lets them through. Until you add it, challenges show as “blocked”, never as down.
Is there a free plan?
Every account gets 650,000 credits on the 1st of every month, forever. That runs 15 monitors every minute, or 75 every 5 minutes, without paying anything. Bought credits last 24 months.
Do you have status pages?
Yes, included. Each page shows only confirmed outages, partial outages and maintenance, never an unconfirmed blip, and carries your name, logo and colours. Each page can run on your own domain, like status.yourcompany.com.
Who runs this?
A small hosting company in Greece that needed it for its own sites, got tired of false alarms, and built the monitor it wanted.
Is there an API?
A scoped public API is on the roadmap after launch. Webhooks for alerts are there from day one.
Be paged only when it's real.
650,000 free credits every month. No card needed.