Introducing Watches: agents that never go off shift
Seven always-on agents that track the slow trends nobody is paid to look at, from filling disks to expiring certificates, and draft the fix before they turn into incidents.
Some of the worst incidents start as a slow trend that nobody was paid to look at. A disk fills by two percent a day. A certificate quietly approaches its expiry date. A third-party API gets a little slower every week. Today we are launching Watches: seven always-on agents that look at those trends for you.
Seven watches, one queue
- Latency checks p95 and p99 on your critical paths every 60 seconds.
- Error budget tracks SLO burn rates every five minutes.
- Deploys compares every release against its baseline.
- Capacity forecasts disks, memory and connection pools 14 days out.
- Cost finds idle and oversized resources every hour.
- Dependencies watches upstream APIs and vendor status pages.
- Certificates tracks expiring certificates and rotated keys daily.
Every finding lands in the same priority queue as your incidents, ranked by the impact Forge expects if nothing changes. A disk that fills in six days ranks above a disk that fills in sixty, even if both are at 80 percent today.
Findings come with a fix
A watch does not stop at a chart. When Capacity predicts that orders-db will reach 90 percent in six days, it drafts the change to expand the volume, explains the growth it measured, and waits for an engineer to approve it. Nothing ships without review.
Watches are available today on the Scale and Enterprise plans, and every team on Team gets Latency, Error budget, Deploys and Certificates. You can also describe your own watch in plain English, and Forge will turn it into a schedule and a set of checks you can read before it runs.