Good tools, different questions.
There are good monitoring tools, and we've used several. None asked the questions we needed answered about our sites, so we built a monitor that does, and only that.
The kinds of tools, and what each is good at
| Kind of tool | Answers | Good at | Needs |
|---|---|---|---|
| Uptime checkers | Is it up, every minute? | Fast alerts, many locations, low cost | Nothing from the site |
| Scripted browser tests | Does this exact flow still work? | Catching a sign-up or checkout that stopped working | A script per flow, kept up to date |
| Real-user monitoring | How fast is it for actual visitors? | The truth about real devices and networks | A script on every page |
| Hosted status pages | How do we tell people about an outage? | Communicating and subscriptions | Someone to post updates |
| Error tracking and logs | Why did it break? | Root causes and stack traces | Access to the code or hosting |
Most teams use two or three of these together, and that's sensible. Each answers its own question well.
What none of them covered for us
- Does it still work for AI agents?
- Many of our sites are built for agents. We check llms.txt, agent cards, OpenAPI and MCP, notice when an MCP server's tools change, and send an agent in with only the address and a goal. Ready for agents →
- Is the newest content actually there?
- The newest stories from each sitemap are checked like pages, and a sitemap that stops updating is noticed. A homepage can be up while publishing has quietly stopped.
- Is anything quietly wrong?
- Pages accidentally hidden from search, a canonical pointing elsewhere, missing security headers, links that lead nowhere, and domains and certificates about to expire, all in one place.
- From the outside, on any host
- No script on the site and no access to its code, logs or hosting. It works the same for a site on any host with any database, which matters as we start watching sites we don't run.
And it reports the way we'd want to read it: numbers rounded for reading, never in our favour, and "since" instead of claiming 90 clean days.
What we don't do
- Passwords. Teams sign in with a code we email; there's nothing to forget or leak.
- Many locations. Checks run from the US West Coast, and a failure is checked again from the US East Coast before it alerts.
- Paging and on-call. Alerts go to Discord, Slack, email or the browser, with no escalation rotations.
- A script on visitors' pages. Readers' speed comes from Chrome's public data instead.
- Open-source the whole service. The checks are open source (below); the history, the second location, journeys and teams around them are the hosted part.
If you need monitoring for your own sites
You can sign up and add your own sites: we email a code, you name your team, and the first check runs within a minute.
Or run ours yourself. agent-ready-uptime is the checks this service runs on, with a small runner around them: uptime, TLS and domain expiry, sitemaps, llms.txt and the other agent files, AI crawler access and MCP tool changes, with alerts to Discord or Slack and a status page. It checks from one place and keeps a day; MIT licence. Run it with npx agent-ready-uptime, Docker or its Railway template.
Other well-known open-source monitors, in alphabetical order, all three active on the date below:
- Gatus
- A developer-oriented status page with health checks set in a config file, and alerting. Apache-2.0 licence.
- Upptime
- An uptime monitor and status page that runs on GitHub Actions, with no server of your own. MIT licence.
- Uptime Kuma
- A self-hosted monitor with a web dashboard and status pages. MIT licence.
For checks that agents care about, anyone can run our free agentic search audit (the open-source readiness-audit) against their own site today.
Last updated .