Introduction

Webmonit is monitoring built for web hosting: it watches machines and the sites they serve, together, and keeps the history of both.

Two programs:

The server (webmonit) — one binary that is the whole service: the scheduler, the checks, the store, the alerting, the web interface, the public status pages and the API. It runs on one machine of yours, with embedded SQLite files beside it. Everything else — the CLI verbs, scripts, panels — is a client of its API.
The agent (webmonit-agent) — one optional binary per watched machine. It reads locally and pushes to the server: metrics every second, CPU and memory per system account, disks, RAID arrays, temperatures, and the sites the machine serves (discovered from the panel or the web server). The server never connects to the agent and holds no credential over it; there is no downward channel by design.

The words

A host is a monitored machine. A site is what a domain serves; it lives on a host, or on none (something external you watch).
A check probes one thing — HTTP, DNS, ping, a port, a certificate — and produces results. A monitor carries the thresholds and the alerting for a check.
A period of failure is an incident; it opens, carries evidence, and resolves. A change that is not a failure — a DNS record moved, a certificate renewed — is an event.
The admin runs the server. Customers are the people whose sites you host; they can be given read-only accounts scoped to what is theirs.

Install the server, then add the first machine and the first site. Alerting and retention have their own pages.