Evidence path
Agents can call tools, connection_exec, skg_search, infrastructure context, and workflow checks to narrow the failing service or dependency.
Root cause analysis in OpsRabbit is not a single model answer. It is a controlled investigation loop that gathers evidence from connected tools, searches service knowledge, follows infrastructure context, tracks hypotheses, and produces a confidence-aware diagnosis with next steps.

What this page covers
Start with scoped evidence from alerts, recent changes, logs, metrics, traces, tickets, deployment history, and SKG hits.
Use connections and MCP tools through approval and sandbox boundaries instead of asking responders to paste raw context manually.
Produce likely cause, impact, missing evidence, validation commands, rollback or mitigation options, and escalation notes.
Operational detail
Root cause analysis in OpsRabbit is not a single model answer. It is a controlled investigation loop that gathers evidence from connected tools, searches service knowledge, follows infrastructure context, tracks hypotheses, and produces a confidence-aware diagnosis with next steps.
Agents can call tools, connection_exec, skg_search, infrastructure context, and workflow checks to narrow the failing service or dependency.
The response should separate confirmed evidence from hypotheses, name missing data, and recommend bounded follow-up checks.
A good RCA thread ends with validation steps, customer-impact language, remediation options, and reusable post-incident evidence.
Related pages