How to Set Up Agent Alert Rules
Create alert rules that notify you when AI agents hit consecutive failures, duration limits, missed runs, or error rate thresholds — via email, Slack, or in-platform.
Agent alert rules let you tell TrustPager what "wrong" looks like for your AI agents — and what to do when it happens. Head to https://app.trustpager.com/auto/agent-alerts to view your alert history and manage the rules that generate notifications.
What this page does
The Alerts page has two tabs: Alert History shows every alert that has fired, split into unacknowledged and already-reviewed. Alert Rules shows the set of rules currently watching your agents. Rules run continuously in the background — you set them up once and TrustPager checks every agent run against them automatically.
The four rule types
When you create a rule via the API or MCP, you choose one of four types based on what you want to catch.
- Consecutive Failures — fires when an agent fails a certain number of times in a row without a successful run in between. Useful for catching agents that have broken silently and are retrying endlessly.
- Duration Exceeded — fires when a single agent run takes longer than a threshold (in seconds). Use this to catch runaway loops or agents waiting on a slow external service.
- No Run Since — fires when an agent that is supposed to run regularly has not run in a given number of hours. Catches missed cron schedules or suspended agents you forgot about.
- Error Rate — fires when the proportion of failed runs crosses a percentage within a rolling time window. More forgiving than Consecutive Failures — good for agents that occasionally fail but should not be failing most of the time.
Notification channels
Each rule has a notify_channel field that controls where the alert goes. The supported channels are:
- email — sends a notification email to the address set in
notify_target. - slack — posts a message to the Slack channel specified in
notify_target. Requires your Slack integration to be connected at https://app.trustpager.com/settings/integrations. - in-platform — creates an entry in the Alert History tab on this page so the alert is visible to any admin who checks in.
You can leave notify_channel blank — it defaults to Slack.
Scoping a rule to one agent
By default a rule applies to every agent in your workspace. Set the agent_name field to restrict the rule to a specific agent. Leave it empty to watch all agents.
Creating and managing rules
Alert rules are created and updated via the TrustPager API or the MCP server. The relevant tools are:
create_agent_alert_rule— create a new rule with your chosen type, threshold, channel, and target.update_agent_alert_rule— change the threshold, channel, target, or enable/disable an existing rule.list_agent_alert_rules— list all rules currently configured for your workspace.delete_agent_alert_rule— remove a rule permanently.list_agent_alerts— read fired alerts programmatically.acknowledge_agent_alert— mark an alert as reviewed without opening the portal.
Rules management is not yet available through the portal UI — use the API, MCP, or ask your AI agent to configure rules on your behalf.
Acknowledging alerts
When an alert fires, it appears in the Unacknowledged section at the top of the Alert History tab with a red banner. Click Acknowledge on any alert to mark it as reviewed and move it to the History section. The alert records who acknowledged it and when, so you have an audit trail if something comes back later.
If everything is running cleanly, the page shows a green checkmark and "No alerts fired" — that is the state you want to see.
Good habit: Create a "No Run Since" rule for any scheduled agent that should run at least daily. It is the simplest way to catch a missed cron or a silently paused agent before it affects your clients.