PigeonAtlas

Email for agents

Let an agent send email. Not to strangers.

Giving an agent an email address is the easy part — three providers shipped it this year. The hard part is that an agent will cheerfully write to a list it found, and every one of those messages is a spam complaint against your domain, in the same reputation your password resets depend on.

We are building the other half: sending an agent can be trusted with. A key that may only send, from one domain, to addresses that asked; a ceiling it cannot cross however the loop goes; and a record, per message, of which agent sent it and why that address was allowed.

What we mean by "asked"

Replied to you

The address wrote to your application first, and the agent answers in the same thread. The safest send there is.

Signed up

A person entered the address in your product and confirmed it. You hold the record of when and from where.

Bought something, booked something, asked for a code

A transaction the person started. This is ordinary transactional mail, and it is what the agent should be sending.

Appeared in a list the agent assembled

Not asked. This is the one an agent will do by default if nothing stops it, and the one that costs you the domain.

What already works today

None of this is waiting on the waiting list. If you are giving an agent send access this week:

The full argument is in the post, including the prompt-injection case that turns a support agent into a mailing list.

Tell me when it is ready

One message when there is something to try. Double opt-in, so the address is not on the list until you click the link we send. No marketing, no sharing, one-click unsubscribe.

Email me when agent sending is ready to try. No marketing, no list sharing, unsubscribe in one click.

Not waiting? The free plan sends 1,000 messages a month from your own verified domain, with the suppression list and the separate queue already doing their job.