Dashboard
Autonomous incident response powered by Gemini + Phoenix
Start an Investigation
The agent handles everything autonomously — just choose a mode.
No investigations yet
Start one above
Waiting
Events appear here in real-time
Active Investigation
—
Integrations
Configure tools. Changes apply immediately.
Cases
All investigations — click to view details
Loading cases...
Live Traces
Real-time Phoenix trace spans from active investigations
No active traces
Start an investigation to see traces flow in real-time
All agent tool calls, LLM reasoning, and guardrail evaluations are traced through Phoenix. In memory mode, traces are collected in-process. In cloud mode, they stream to Arize Phoenix for persistent analysis.
Accuracy
Guardrail pipeline performance — hallucination detection rates
Every finding produced by the AEGIS-IR agent passes through the Guardrail Pipeline before reaching analysts:
APPROVE Finding is backed by evidence — passed to analyst.
FLAG Finding needs human review — evidence is partial.
BLOCK Finding is hallucinated — blocked automatically.
The agent learns from blocked findings via the Self-Improvement Loop, adjusting its behavior to reduce hallucinations over time. All evaluations are traced through Arize Phoenix for observability.