Policies
Declarative rules that govern agent actions.
Policies contain an array of rules. Each rule has an if condition (field, operator, value) and a then action. Policies can be scoped to organization, workspace, agent, environment, or global.
Policy types
- data_protection, prompt_restrictions, api_restrictions, tool_permissions
- human_approval, logging, retention, geography, custom
Rule structure
json
{
"id": "rule-1",
"if": { "field": "action", "operator": "equals", "value": "export" },
"then": { "action": "block", "message": "Exports are not allowed." }
}Operators
- equals, not_equals, contains, not_contains
- matches (regex), greater_than, less_than