Duskwatch tells on-call engineers what broke, and whether it can wait

DigitalOcean sends its alerts by email or Slack, and spend alerts by email only. Acting on a problem usually means opening a laptop. For the person on call, the first question is simple and hard to answer from a phone: what broke, and can it wait until morning?

Answering it means piecing together graphs, deploy activity and the status page, and half of that is ruling out the platform itself. During a platform-wide incident, alerts arrive one after another for problems nobody on the team can fix.

There is also a trust problem. An app that can reboot servers needs powerful access to your cloud account. If that access leaks, or sits on someone else's server, it becomes a risk.

We designed for the moment the alert lands: four steps, then back to sleep. Everything else waits for the laptop. Duskwatch does not replace DigitalOcean's alerts or control panel, and it never creates, resizes or deletes anything.

Access follows least privilege. Signing in only lets the app look. Changing things is a second permission you approve on its own, and it never leaves the iPhone. Alerts use a third, read-only permission held by a small server in Frankfurt, which rejects anything that can write.

Privacy is built into the design, not written into a policy. Notifications carry only IDs and the phone fills in the names, so the server never learns what your servers are called. Findings are plain sentences, worst first, and a likely cause is labeled as a guess.