For agencies
An ops team monitors one product it knows intimately. An agency monitors forty sites it built at different times, on hosting it did not always choose, for clients who each pay differently and each expect to be told first. Same word, monitoring. Different job. Kockpit is built for the second one.
Group sites the way you actually think: by client, by the person responsible, by category, by hosting account. Whoever is attached to a tag hears about those sites, and nobody hears about the rest.
This is the difference between an alert and noise. When everything goes to one channel, monitoring becomes something the whole team learns to ignore, and then it might as well not exist. When the designer who owns three retainer clients hears about exactly those three, the alert still means something at month twelve.
Every client's retainer sits beside their sites: what they pay, what hours are included, what has been logged this month, and what is collectible. Work gets logged against the client, and the allowance counts down where everyone can see it.
Quiet over-servicing is how agency margin disappears. Fifteen minutes here, a small favour there, none of it billed, all of it invisible until someone counts. Kockpit counts. When a client is at eighteen of twenty hours on day ten, that is a conversation about scope, held early, instead of a write-off discovered at invoice time.
Before a site goes live, Kockpit runs the list everyone keeps meaning to write down: the HTTPS redirect, the favicon, a real 404 page, leftover placeholder copy, the copyright year, a reachable contact route, protection on the forms.
None of these is hard. All of them are embarrassing in the client's launch email. A checklist that runs itself is the difference between a launch process and a launch tradition.
One button turns any scan into plain text: the domain, then a bullet per fixable issue. It goes straight into the ticket system, the client email or the standup note. Nothing to reformat, no screenshots of a dashboard, no login for the subcontractor who just needs the list.
Every morning Kockpit emails its own health: whether the checks ran, whether the digest drained, whether every site has someone to alert. It sends even when everything is fine, on purpose.
A monitor that has quietly died looks exactly like a month with no problems. For an agency selling "we are watching your site" as a line item, that failure mode is not hypothetical, it is a refund. If the heartbeat stops arriving, that silence is the alert.
Honestly
If you run one product on infrastructure you own, with status pages, on-call rotations and sub-minute checks, you want an ops tool. Kockpit does not do PagerDuty-grade incident management and is not trying to.
Kockpit is for the agency and the freelancer with a folder of client sites, a spreadsheet of renewal dates, and a monthly line item that says "maintenance". It replaces the spreadsheet, and it makes the line item defensible. What it watches is documented in detail here.