Alerts and Issues
Alerting is the spine of AgentPing. You set rules for what matters; we tell you once when something breaks and once when it recovers, and we never spam you in between.
What you can be alerted about
Each rule watches for one thing, scoped to a single agent or a whole team:
- An agent goes quiet. It stopped reporting inside its expected window. Every account starts with one of these by default.
- Runs start failing. A run failed, or several failed in a row.
- Health drops. The share of healthy runs falls below your target over a recent window.
- Quality drops. The average evaluation score over recent runs falls below your target.
- Spending runs hot. A budget threshold is crossed, or spending is on pace to exceed a budget.
You choose which channels each rule notifies and, optionally, how often to remind you while an issue stays open.
One notification when it breaks, one when it recovers
When a rule's condition becomes true, AgentPing opens an issue and notifies your channels once. While that issue is open the condition does not re-fire; you do not get paged every cycle. When the condition clears, the issue resolves and we send a single recovery note. If you set a reminder, an unacknowledged issue is nudged at that cadence, and nothing more.
We page you. We don't spam you.
The Issues page
Issues in the sidebar is the single place that answers "what is wrong across my agents right now, and what have I dealt with?" The nav badge shows how many issues are open.
- Open lists everything still live, worst first. Issues on the same agent collapse into one group, so a struggling agent reads as one problem with a count, not a wall of duplicates.
- Resolved shows the last 30 days, read-only.
Acknowledge and snooze
On any issue, or straight from the notification's one-tap link:
- Acknowledge stops reminders but keeps the issue open until the condition actually clears.
- Snooze (1 hour, 24 hours, or 7 days) goes quiet for that rule and target for the duration, and resolves the current issue.
- Resolve closes it now.
Acknowledge a group to silence every open issue on one agent at once.
Resolved issues are kept for 30 days.