Skip to main content
A Brain’s rules shape what your agents do, so two rights are kept apart: drafting a rule and putting it into force. Both are ordinary role permissions under Settings → Users → Roles.

​
The two rights

Reading a Brain — browsing rules, the graph, history, and asking it questions — needs only View Brains.

​
Two typical roles

  • Approver — Edit Brains + Approve Brain Rules. Domain owners and ops leads who decide what the agents follow.
  • Contributor — Edit Brains only. Analysts and subject-matter experts who propose rules. Everything they add waits in Needs review until an approver puts it into force; the approve, retire and restore controls are not shown to them, and the API answers 403 if called directly.
Org owners, global admins and workspace admins hold both rights.

​
What changes for existing accounts

Nothing, until you decide it should. Every role that could edit rules before this permission existed was granted Approve Brain Rules automatically. To create a contributor role, edit the role and untick Approve Brain Rules.

​
Approving your own draft

Allowed. Approval is a right, not a second pair of eyes — an approver who drafts a rule can put it into force in the same sitting.

​
Auto-approval

Two features file rules into the graph without a per-rule click, and both are approver-only:
  • “Approve automatically” on an import, upload or repository build. Without the approve right the request is refused rather than quietly downgraded, so a script is never told its rules went live when they did not.
  • Auto-approve recurring rules in the Brain’s Settings, which lets a draft that keeps being proposed from real tasks approve itself. Only an approver can turn it on or off.

​
Agents and Brains

An agent reads the one Brain configured in its settings. The get_brain_context tool follows the same rule: with a Brain configured, it only ever consults that Brain.