StatusCake

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Build me the monitoring that replaces StatusCake — and note first that **its free tier covers a small site**, so the honest answer for a personal project is to use it.

Build your own when you want the history in your own database, or when you are already paying.

STACK
- Node 20+ with Fastify
- SQLite through better-sqlite3, WAL mode
- A worker on a server elsewhere, ideally two in different countries
- Caddy in front

THE FOUR CHECKS THIS CATEGORY ACTUALLY SELLS

UPTIME
- HTTP with a timeout, asserting status and a body string. Record every result with the timing breakdown
- Confirm failures from a second location before opening an incident
- One incident per check, alert on open and on recovery

PAGE SPEED
- Lighthouse on a schedule against your key pages, storing the full result
- **Track the metrics that matter — largest contentful paint, cumulative layout shift, interaction latency — as a series, and alert on regression rather than on an absolute number.** Absolute thresholds cause alert fatigue; a page that got twice as slow last Tuesday is a real finding
- Store the run so a regression can be diagnosed rather than merely noticed

DOMAIN EXPIRY
- WHOIS or RDAP on each domain weekly, alerting at ninety, thirty and seven days
- **This is the cheapest catastrophe to prevent.** A lapsed domain takes down email as well as the site, and recovering one after it is bought is expensive or impossible

CERTIFICATE EXPIRY
- Fetch the certificate, read the expiry, alert thirty days ahead. Also check the chain and the hostname match, because a certificate that is valid but for the wrong name is a real failure mode

THE DATA MODEL
- checks, results, incidents, maintenance as usual
- assets: id, kind, value, expires_at, checked_at — domains and certificates share this shape

ALERTING
- Deduplicated, escalating if unacknowledged, quiet hours with a severity override, and always a recovery message
- Maintenance windows recorded rather than hiding the history

THE STATUS PAGE
- Static, on a different host and domain from the application it reports on
- Honest uptime with maintenance shown separately, and ninety days of history

WHAT MATTERS MOST
Domain and certificate expiry, and regression-based speed alerts. The first two prevent the outages nobody sees coming; the third is the only speed alerting anybody keeps switched on.