AI security
AgentGuard
Supervise autonomous agents in real time. AgentGuard flags prompt injections, sensitive file access, risky outbound transfers, suspicious requests, unauthorized tools, and policy violations—so you can approve, block, and audit before damage lands.
Open Dashboard (demo)
Set AUTH0_* and SECRET_KEY in .env to enable login.
Live supervision
Streaming activity feed with severity and policy context for every flagged step.
Review queue
Pending actions surface Approve and Block controls with a full rationale panel.
Session audit
Allowed vs blocked counts, pending review, and an exportable JSON report for demos.