Exact Action Name, category prefix, or "*" wildcard pattern
(e.g. "Send Single Message", "*Email*", "Place Outbound*").
OptionalconditionOptional SafeExpressionEvaluator boolean condition evaluated against
{ params, user, now } (e.g. "params.Amount > 500 || params.IsExternal === true").
When omitted, the rule applies unconditionally to matching actions.
Execution gate applied when this rule matches:
'Allow': Proceed automatically (within the user's permission ceiling)'RequireApproval': Pause and create an MJ: AI Agent Requests approval item'Deny': Block execution and return a policy denial message to the agentOptionaltimeOptional time-of-day / day-of-week windows restricting when the action may run (e.g. allowing outbound follow-up nudges only Monday–Friday 08:00–18:00).
OptionaldaysOfWeek?: ("Sun" | "Mon" | "Tue" | "Wed" | "Thu" | "Fri" | "Sat")[]Allowed days of week (e.g. ['Mon', 'Tue', 'Wed', 'Thu', 'Fri']).
OptionalendTime?: string24-hour end time in HH:mm format (e.g. "18:00").
OptionaloutsideWindowBehavior?: "ReadOnly" | "Deny" | "RequireApproval" | "DeferUntilNextWindow"Behavior when the action is invoked outside the allowed window:
'Deny': Reject immediately'RequireApproval': Create an AIAgentRequest for user approval'DeferUntilNextWindow': Return nextWindowStartISO so the agent can self-schedule an AgentWakeOptionalstartTime?: string24-hour start time in HH:mm format (e.g. "08:00").
Optionaltimezone?: stringIANA timezone (e.g. "America/Chicago"); defaults to UTC when omitted.
OptionalvalidFromISO?: stringOptional ISO-8601 start timestamp for temporary grants.
OptionalvalidUntilISO?: stringOptional ISO-8601 end timestamp for temporary grants.
Per-action or wildcard action policy rule stored in the
ActionPolicyRulesJSON array onMJ: AI Agent User Policies(AIAgentUserPolicy.ActionPolicyRules).Evaluated by
BaseAgentbefore executing any action or tool on behalf of the user.