updown.io

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 minimal uptime monitor that replaces updown.io — with the honest note that **at a few euros a month it is barely worth replacing.** Build it if you want the history in your own database.

The attraction of the original is that it does one thing. Keep that.

STACK
- Node 20+ as a single small process
- SQLite through better-sqlite3, WAL mode
- Run it on a server that is not the one being watched
- Caddy in front, only for the status page

THE WHOLE THING
- checks: id, url, interval_s, timeout_ms, expect_status, expect_body, is_active
- results: id, check_id, at, ok, status, ms, error
- incidents: id, check_id, opened_at, closed_at
- That is the schema. Resist adding to it

THE LOOP
- Every interval, fetch with a timeout, assert status and body, write a result
- Two consecutive failures open an incident; two consecutive successes close it. **The two-in-a-row rule is what stops a single network hiccup waking you** and it is the only debouncing you need
- Record the timing breakdown — resolve, connect, TLS, first byte — because it costs nothing and it is what tells you where 'slow' lives

APDEX, WHICH IS THE IDEA WORTH COPYING
- Pick a target response time. A response under it is satisfied; under four times it is tolerating; above is frustrated
- Apdex is satisfied plus half the tolerating, over the total. **One number between 0 and 1 that means something**, computed at query time from the results
- It beats an average, which hides the bad tail, and it beats raw uptime, which says a site crawling along at eight seconds is fine

ALERTING
- Email and a push notification. Two channels is enough
- One message on open, one on recovery, with the duration. Nothing else
- A daily digest only if there was something to say

THE STATUS PAGE
- Static HTML regenerated after each check round, hosted somewhere other than the application
- The last ninety days as a row of bars per check, response time as a line, and the current Apdex
- **No JavaScript required to read it.** A status page is read when things are broken, sometimes on a bad connection

WHAT TO REFUSE TO ADD
- Dashboards, teams, integrations, an API. Every one of those turns a tool you never think about into one you maintain. The whole appeal is that it does one thing

WHAT MATTERS MOST
The two-in-a-row rule and Apdex. One keeps the alerts trustworthy; the other gives you a single honest number about how the site actually feels.