A

audit-log

Authorization and execution control for AI agents — independent policy and judge approval before consequential tool actions execute. MCP gateway, capability tokens, auditability, and fail-closed execution.

Stratum is an AI governance layer for MCP systems that sits between model intent and tool execution. It enforces permissions, policies, context controls, approval workflows, sandboxed execution, and audit logging to ensure every action taken by an AI system is authorized, traceable, and constrained by explicit rules before it runs. https://roxanneardary.com/stratum/