trust-model
Runtime security guardrails for AI agents — inspect, control, and audit every tool call. Zero-dependency local mode included.