Governance

Complete operator oversight without requiring any operator involvement in normal resolution. Live supervision surfaces active conversations with per-conversation sentiment, confidence, and tier state via SSE. Prompt A/B testing runs with statistical significance tracking so you know before you promote a new version.

• ✓Real-time supervision: live feed of all active conversations across every channel
• ✓Agent run replay: re-execute any conversation against any model version
• ✓Prompt A/B testing with statistical significance tracking and one-click rollback
• ✓OpenTelemetry export for Datadog, Grafana, Honeycomb, any OTEL backend