Uptime Kuma is excellent. Sometimes it's still the wrong tool.

Kuma has a single admin account, everyone who can see the dashboard can silently change or delete anything, and there's no trail of who did what. Vigil ships unlimited team members across four roles (owner, admin, responder, viewer) and records every incident state change: who, what, when. There is no seat cap and no per-user price.

Serving clients from Kuma means one shared instance where everything sees everything. Vigil is multi-tenant by design: each client is an isolated organization with its own monitors, members and a branded public status page on your domain. One install, a page per client. This is the agency setup, and it's covered by one license.

Kuma tells you a check failed. Vigil opens an incident with severity and a timeline, investigating → identified → monitoring → resolved. Publishes it to the status page, closes it on recovery, and drafts the postmortem from the real events (AI, on your own key, optional). When a client asks what happened, the timeline is the answer.

Kuma runs its checks from the instance it is installed on, and customer-run probe agents with a configurable quorum are not in its feature list. So "the target is down" and "the path from this one box is down" arrive as the same alert at the same hour of the night. Vigil can dispatch a monitor to probe agents you run on your own machines and require several to agree first. A probe that does not answer is never counted as a failure. All four outcomes, recorded from a real run, and how to set it up. Your machines: Vigil hosts no regions and charges nothing per probe.

The most complete free alternative: uptime, incidents, status pages and on-call, open source. It is a much heavier platform to operate, where Vigil is two processes and one Postgres on a $5 VPS, with a commercial license and one accountable human behind support. If you have the ops capacity for a big platform, evaluate it.

Then take Vigil Core, the free edition under Apache-2.0: github.com/sikurdev/vigil-core. HTTP checks with keyword assertions, incidents, a public status page, team roles, and every check type the paid edition has. Both are cut from one commit. Same two processes. It exists so you can judge the engineering before you decide, and it is a perfectly good place to stop.