Docs/Aigovernance/Airesilient/Alerting

Alerting & Notifications

ANTS Platform sends notifications for specific product events and hands them to the tools your team already uses: Slack, PagerDuty, ServiceNow, Jira and webhooks.

⚠

The platform does not alert on latency, error-rate or cost thresholds of your traces. For threshold alerts, poll the Metrics API from your own monitoring, as shown below, or use Budget AI hard budgets to stop spend.

What can notify you

Source	Event	Destinations	Docs
Automations	Any event in the automation catalog, such as guardrail violations, tool misuse, drift alerts, agent suspension, Discover AI and device events, budget thresholds, key expiry, exports and prompt versions, with AND/OR conditions	Webhook, Slack, Microsoft Teams, PagerDuty, ServiceNow, Jira, Linear, email, in-app	Automations
Governance	An unapproved model is used, or an agent drifts from its governance profile	Slack, Jira, ServiceNow incident, PagerDuty	Drift Alerts
AI Optimizer	All evaluations for a comparison model failed; a cheaper model saves more than 10%; a faster model cuts p95 latency by more than 20%	ServiceNow risk record	Recommendations
Budget AI	Spend crosses a budget's alert threshold	Recorded under Budget AI → Budgets & Limits as threshold crossings. To send a message, create an automation rule on the budget threshold event	Budgets & Limits

Connect the destinations under Settings → Integrations. See Slack, PagerDuty, ServiceNow and Jira. ServiceNow risk records also need Enable Risk Records turned on in the ServiceNow settings.

Threshold alerts with the Metrics API

Query the metric you care about on a schedule from your monitoring system and alert when it crosses your threshold. This example returns p95 generation latency per model for one hour:

bash

curl -G "https://<your-host>/api/public/metrics" \ -u "pk-ap-...:sk-ap-..." \ --data-urlencode 'query={"view":"observations","dimensions":[{"field":"providedModelName"}],"metrics":[{"measure":"latency","aggregation":"p95"}],"filters":[],"timeDimension":null,"fromTimestamp":"2026-10-18T09:00:00Z","toTimestamp":"2026-10-18T10:00:00Z","orderBy":null}'

Use {"measure":"count","aggregation":"count"} grouped by level for an error share, or totalCost with sum for spend. See Metrics & Monitoring for all measures and the API Reference for authentication.

Next steps

Incident Response: what to do when an alert fires.
Metrics & Monitoring: the measures behind your alerts.