whitepact

WhitePact is an open-source runtime authority and governance layer for AI agents and autonomous systems. It sits between agent intent and consequential execution, evaluating identity, authority, intent, policy, risk and approval before an action is allowed to proceed. WhitePact supports explicit outcomes including ALLOW, ALLOW WITH REDACTION, REQUIRE APPROVAL, DENY and QUARANTINE, together with evidence and audit support. It is designed for agentic systems, MCP-based workflows and enterprise environments where authentication alone is not enough to establish whether a specific action should execute. WhitePact is currently in qualification-stage development and is being hardened before broad production launch.