HTTP and HTTPS
The basic one: does it answer, how fast, and with what status code. Redirects are followed and each hop is checked.
Eight ways to find out first
A site can return 200 OK and still be broken. Most tools stop at the status code. These do not.
The checks
Keyword match
Look for a word that should be on the page, or one that should not. A WordPress site showing "Error establishing a database connection" answers with a 200, and this is what catches it.
JSON assertion
For APIs. Point at a field in the response and say what it should be. An endpoint that returns 200 with an empty payload is down as far as whoever depends on it is concerned.
SSL certificate expiry
Warns before the certificate lapses, not after. An expiring certificate is reported as degraded rather than down, because the site is still serving perfectly and nobody should be paged at 3am for a nine-day deadline.
Domain expiry
The one that ends agencies. Answered over RDAP rather than WHOIS, so it works on shared hosting where port 43 is blocked. A few country domains publish no RDAP at all — those report that they could not be checked rather than pretending they passed.
DNS records
Assert that a record still resolves to what it should. Catches a registrar change or a nameserver edit before the traffic does.
TCP port
Is the mail server, the database or the SSH daemon accepting connections. Not everything that matters speaks HTTP.
Heartbeat
The inverse: your cron job pings a URL when it finishes, and silence is the alarm. This is how you find out a nightly backup stopped running, which nothing else on this list can tell you.
It knows what a broken site looks like
A Laravel stack trace, a WordPress critical-error screen, a LiteSpeed 508 resource limit. These come back with a perfectly healthy status code and a page your client cannot use. The signature list is maintained by us and applies to every account.
Told once, properly
One outage produces one alert thread, not a message every minute until somebody mutes the channel. Reminders are off by default and escalation is on, because a tool that repeats itself is a tool people stop reading.
Where alerts goWhat it does not do yet
Checks run from one place, so a network problem between us and your client looks the same as their site being down. There is no ICMP ping and no sub-minute interval. All three need probe servers in more than one country, which is the next piece of infrastructure rather than a feature we are hiding behind a plan.
Set it up this afternoon
Add your first monitor in about a minute. Add the rest of the portfolio over a coffee.