Your monitoring shouldn't die
with the server it watches.
MT-Uptime checks your sites, servers, databases and TLS certificates from outside your infrastructure — and tells you what actually broke, not twenty times over.
No card required · cancel anytime · the whole engine is open source
The monitor that never becomes the outage
Most monitoring tools poll harder when a target starts struggling. A failed check fires an immediate retry, retries stack up, and checks that haven't finished overlap with the next ones. That is how a monitor turns a limping server into a dead one, at exactly the wrong moment.
MT-Uptime is built so it cannot do that:
- A check never outlives its interval. The timeout is clamped below the interval, so two checks of one monitor can never be in flight at once.
- Failure never speeds up polling. Retries happen within the existing schedule, not in addition to it. A target that is down gets exactly the same request rate as one that is up.
- One request in flight per target, always. No thundering herd the moment something recovers.
- Defaults that respect the answer's shelf life. Certificates are checked every 6 hours and DNS every hour, because those answers change on the order of days.
One server died. Not twenty sites.
When a host carrying twenty monitored sites goes down, every other tool sends twenty alerts and leaves you to work out that it was one machine. MT-Uptime resolves each monitor to the infrastructure behind it and opens one incident when monitors sharing a host fail in the same window — with the response times, DNS resolution, status codes and certificate state that tell you what broke.
What it watches
HTTP and HTTPS
Status ranges, keyword present or absent, redirects, response time. Basic and Bearer auth, custom headers, request bodies and a per-monitor User-Agent, for endpoints behind a login or a WAF.
TLS certificates
Expiry with a warning window you choose, so a renewal that quietly failed becomes an alert weeks before it becomes an outage.
DNS
A, AAAA, CNAME, MX and TXT records, optionally against a specific resolver, with expected-value matching.
Databases
MySQL and PostgreSQL, checked with a real connection and a real query — not a port knock that passes while the database refuses logins.
TCP ports
Reachability for anything that isn't HTTP: mail, SSH, game servers, message brokers.
Cron and backups
The inverse check: your job calls us, and we alert when the call doesn't arrive. A dead-man's switch for backups and scheduled tasks.
And what it does about it
Alerts that aren't noise
A single failed check doesn't page you — you choose how many consecutive failures confirm an outage. A definitive negative skips the wait, because retrying won't change the answer.
Slow is a problem too
A 200 that takes eight seconds is still broken. Set a threshold and get told, after enough consecutive samples that it isn't just noise.
Nine ways to be told
Email, Slack, Discord, Microsoft Teams, Telegram, ntfy, Gotify, PagerDuty or a webhook. PagerDuty gets real incident semantics, so a recovery resolves the incident instead of paging someone about a service that is already back.
Status pages for your customers
Public pages with 30-day uptime per monitor, so you can point people somewhere instead of answering the same question fifteen times.
Maintenance windows
Planned work doesn't page anyone, and doesn't quietly ruin the uptime figure you report to a customer.
Uptime maths that survives pruning
Raw checks are rolled into hourly and daily buckets before they're deleted, so a long-range uptime percentage stays accurate instead of silently drifting.
The hosted version runs the exact code you can download
MT-Uptime is open source under the AGPL. Not a stripped-down community edition — the whole monitoring engine, the one we run. You can read every check we make on your behalf, and if you'd rather run it yourself, you can, for free, forever.
So what are you paying us for? Somewhere else to run it. An uptime monitor has one unfakeable problem: it cannot monitor itself, and it dies with the infrastructure it watches. That is not a gap you can close by installing more software on the same box. It is the entire reason this product is worth hosting, and it is why we don't have to withhold features to sell it.
One plan. $29 a month.
Everything above. No tiers, no feature grid, no surprises.