You set the boundaries. AI works freely inside them. At the edge, it stops or asks.
People write the rules, and each rule records who set it.
When an AI agent wants to do something, it first proposes the action.
A deterministic check, which gives the same answer every time for the same rules, replies yes, no, or wait for a person. It names the rule that decided and who set it.
Nothing runs until that answer comes back.
For developers
$ npm install @neuroverseos/governance
Documentation lives in the repository.
The 1.0 release candidate, generalised from our robotics rulebook work, is on npm.
Its first application is our robotics project, robot-domain-rules.