Skip to main content
An approval puts a person between an agent and an action. The agent asks, the action waits, and someone with the approver role decides in the dashboard. Bud holds a send for approval when one of your policy rules says it needs a person: for example, mail to a domain outside a list you keep. Sends that no such rule covers go straight out.

Who approves

Approvers are workspace members you mark as approvers. By default, the person who owns the agent is its approver. An agent can never approve its own action.

Deciding

Open Approvals in the dashboard, review the request, and approve or reject it. You must be signed in as an approver. Notification emails link to the dashboard. The link opens the request and never approves or rejects it, so a mail scanner that fetches links can’t decide anything.

What approval covers

An approval authorizes exactly that action, once. If the content changes, the approval no longer applies. If your rules now forbid the action, it is refused even though you approved it. A request nobody decides expires, after 24 hours by default. An expired or rejected action is not carried out.

What the agent sees

A held send answers with state: held and an approval_id. The message goes out once an approver approves it, and never otherwise. Nothing the agent sends can release it, so it should not send the message again. A held message has used its sending budget; the budget is given back if the message is rejected or the request expires. Where an action can’t be held, it is refused with approval_required. See Errors.