Last released Oct 10, 2026
Open-source safety layer for AI agent tool calls: policy, risk checks, human approval and tamper-evident audit