03 · ImproveComing soon
Monitor
Catches live issues and triages them for diagnostics
Watches your deployed agents via event and threshold triggers, uptime, latency, cost, eval-score drift, and online evals on live traffic. Detects anomalies, routes them to the right channel, and tees them up for the Diagnostics Agent.
Monitoring / Incident Agent is the entry point for the Improve loop. A regression in tool selection, an eval drift past your threshold, an anomaly in latency or cost: it catches the signal, posts it to the right Slack channel, and hands it off cleanly to Diagnostics.
What it does
- Tracks uptime, latency, cost, and eval-score drift
- Runs online evals on live traffic
- Fires on thresholds and anomalies, routes the incident
Inputs
- Live production telemetry
- Eval thresholds
- Routing policies
Outputs
- Always-on monitoring on your live traces
- Dashboards, drift detection, anomaly alerts
- Incident hand-off into diagnostics
Works with
Datadog
OpenTelemetry
Slack
GitHubGet early access to Monitor
Join the early-access list and we will reach out the moment this agent ships.