It automatically determines why an incident happened — joining metrics, logs, traces, events, and topology, and using causal AI to trace a symptom back to its originating fault, with evidence and a recommended next step.

Correlation surfaces anomalies that coincided and leaves you to guess which mattered. Causal AI traverses your live dependency graph to determine which entity actually caused the failure — causation, not coincidence — and shows the reasoning behind it.

Every root cause comes with an evidence timeline, the supporting signals, and a confidence score explaining how and why the entity was identified. Agentic causal RCA has reported accuracy in the low-80s in demanding Fortune 100 environments — and because it’s explainable, engineers can verify the call quickly.

It builds and maintains a real-time map of your services and their dependencies, then reasons over that structure to localize faults — identifying the originating service even across multiple layers and partial topology.alerts and real deviations still stand out.

Yes. It links incidents to recent deploys and configuration changes, ranks the likely change cause, and can suggest the rollback — since a change is the most common trigger for an incident.

Beyond clustering signals it already has, an investigation agent actively gathers new evidence during an incident — querying logs and metrics, hitting cloud APIs, reading CI/CD pipelines, and traversing the dependency graph — reasoning across multiple steps to reach a root cause.