Investigation workflowVerified
Diagnostic repair: before
Follow a real agent and reviewer through tests, denied diagnostics, evidence search and a two-run comparison.
EXPLORE THE EVIDENCE
Demo library
Follow what an agent did, from a single tool call to a team across Pods.
Ready to exploreSigned captures · checked at build
Recorded executions 9
Open a capture. Follow its evidence.Investigation workflowVerified
Diagnostic repair: after
Continue the same native session, disable unnecessary diagnostics and inspect the successful report.
Single agentVerified
One agent: supply chain
Follow a poisoned install from tool call to file, network and artifact evidence.
Agent teamVerified
An agent team
Follow delegation, peer influence, a refusal and a later install attempt.
KubernetesVerified
Across Kubernetes Pods
Inspect live cross-Pod network and runtime evidence alongside replayed application hooks.
KubernetesVerified
Kubernetes placement
Inspect workload placement and container identity in a captured execution.
Outbound dataVerified
Grok: historical investigation
Inspect a dated outbound-data investigation and its documented evidence limits.
Outbound dataVerified
Grok: three outbound routes
Distinguish model requests, vendor telemetry and third-party analytics in a dated capture.
Single agentVerified
DeepSeek: a development task
Inspect a real task, tool results, saved file changes and partial runtime association.
External evaluators 2
Build on the same execution evidence.LLM as security analyst
Setup requiredAn external evaluator consumes evidence and returns referenced signals; its own requests can be audited.
Read setup guideJev reference integration
Setup requiredCompare typed decisions and manually review them in the separate evaluator workbench.
Read setup guide