Skip to Content
ModulesRCA Lab

RCA Lab

The RCA Lab is AtlasAI’s AI-powered root cause analysis workspace. It combines multi-signal evidence analysis, historical pattern matching via RAG, and topology-aware reasoning to produce ranked root cause hypotheses with explainable evidence chains.

Key Features

  • Multi-signal analysis — Analyzes metrics, logs, traces, alerts, topology, and change events simultaneously
  • RAG-powered reasoning — Retrieves context from past incidents, runbooks, and documentation to inform analysis
  • Topology-aware RCA — Walks service dependency graphs to trace failures upstream and downstream
  • Confidence scoring — Each hypothesis includes a confidence percentage based on evidence strength
  • Evidence chains — Transparent reasoning showing exactly which data points led to each conclusion
  • Comparative analysis — Compare the current incident to historical incidents with similar signatures
  • Interactive exploration — Drill into any hypothesis to see supporting and contradicting evidence
  • Feedback loop — Mark RCA results as correct, partially correct, or incorrect to improve future accuracy

How to Access

Navigate to RESPOND → RCA Lab in the left sidebar to access the standalone lab, or click Run RCA from within any incident (command palette: Cmd+K → “RCA Lab”).

Basic Usage

  1. Open the RCA Lab from the sidebar
  2. Select an incident to analyze, or create a new investigation
  3. Review the auto-collected evidence — the lab pulls available metrics, logs, and topology for the affected service
  4. Add additional evidence manually if the auto-collected data is insufficient
  5. Click Analyze to run the reasoning engine
  6. Review the ranked hypotheses in the Results panel
  7. Click into a hypothesis to see the full evidence chain
  8. Mark the correct root cause to improve future RCA accuracy
CorrelationRunbooks