Trust

Trust is how a person tells a mailbox which recipients are safe to write to. It lives in the mailbox's allow list, and only a person can change it.

The allow list

The allow list holds addresses and domains. An entry for ada@example.com trusts that address. An entry for example.com trusts every address at that domain. The agent can read the list with mail_rules. It cannot add to it.

What trust changes

Trust removes the first contact hold for those recipients, and in filtered mode it lets a message go when every recipient is on the list. It does not override the other holds:

• In review mode every message is held, trusted or not.
• In filtered mode a message goes at once when it is in a conversation a person started, or when every recipient is on the allow list. A message with an attachment, or in a flagged thread, is still held.
• In none mode the first contact hold is the one that trust lifts.

A reply to machine mail, and a reply to a message that failed DMARC alignment, are held whatever the allow list says. See Gate modes.

Always trust

When a person approves a held message in Review, they can choose Always trust… instead of plain approve. It lists each recipient, and the person ticks the ones to add. The message is sent, and the ticked addresses join the allow list. The list holds up to 200 entries.

On an agent's own mailbox, anyone who may decide can trust. On a shared mailbox only a workspace owner can. The mailbox's activity records each change as "Trusted a contact" or "Stopped trusting a contact".

Trust and screening

Trust covers who the agent may write to. Screening covers what the agent may read. A trusted contact's message is still screened on arrival, and a flagged one is still held back from the agent until a person releases it.