A policy layer
A place to decide whether a tool action should be allowed, denied, approved, or dry-run.
About
RuleOak makes AI agent actions governable
The project focuses on the moment where AI moves from answer generation into real-world action: tools, files, shell commands, emails, database writes, deployments, connectors, and MCP calls.
What RuleOak is
An evidence layer
A structured record of what the agent attempted and why the action was permitted or stopped.
A developer layer
A small runtime and CLI that developers can adopt without replacing their agent stack.