Alerts

Get notified when a service crosses a threshold — and jump straight to the cause.

Rules

Define rules on a metric (error rate, errors, requests, average latency, total tokens), an operator (>, ≥, <, ≤), a threshold, and a scope (any service, a specific agent, or a specific MCP tool). Rules are evaluated live against the summary data. A rule that fires ≥2× past its threshold is flagged critical; otherwise warning.

Rules are stored client-side (localStorage) — no backend configuration needed.

Jump to the error

Every firing alert has a View failing traces action that navigates straight to the offending service's traces, pre-filtered to errors (/traces/<type>/<service>?status=ERROR). Global alerts link to the errors view. This turns "something's wrong" into "here's exactly what's wrong" in one click.

Firing alerts appear both on the Alerts page and as a top-5 card on the Summary.