Every error event or failed span is fingerprinted into an error group as it lands. When errors burst, the incident builder loads the failing traces, finds the deepest failing spans with no failing child, ranks candidate roots by trace share, onset order and symptom structure, and reverses the calls graph from the root project to compute the blast radius — observed affected services and predicted ones. The result is a failure graph in the workspace, an onset timeline, and an agent-written explanation you can hand straight to an incident-response playbook.
Early access — turned on per workspace. Ask us to enable it for yours.
- Local roots ranked across failing traces — the first failure, not the loudest symptom
- Blast radius from the reverse call graph: observed vs predicted affected services
- Failure graph per incident with onset timeline and ranked roots
- Agent explanation: what failed first, how it spread, most likely cause
- Run playbook hand-off to incident-response or auto-root-cause-analysis
180
days of traces & incidents