Evidra Lock
Validates AI agent infrastructure operations against explicit policy before execution, acting as a fail-closed kill-switch with evidence-backed decisions.
Validates AI agent infrastructure operations against explicit policy before execution, acting as a fail-closed kill-switch with evidence-backed decisions.
Evidra-lock functions as a critical kill-switch designed for AI agents interacting with production infrastructure. It intercepts and validates every destructive command from AI agents (like those using `kubectl`, `terraform`, or `argocd`) against explicit, deterministic Open Policy Agent (OPA) policies before execution. The tool operates on a fail-closed principle, denying any operation that is dangerous, incomplete, ambiguous, or otherwise fails to meet policy requirements. All decisions, whether allow or deny, are recorded in an immutable, SHA-256 hash-chained log, providing tamper-evident proof for audit and compliance. It also integrates seamlessly into traditional CI pipelines, applying the same robust validation engine.