How can an AI agent receive a verification email?
Give the agent an inbox it is allowed to read. It waits there for the new message, takes the code or link, and goes back to the sign-up.
A browser alone cannot read a separate mailbox. The agent needs its own way into an inbox: after the service sends its verification message, the agent waits for a new message from that service, takes the code or link, and returns to the sign-up flow.
A builder discussion describes this interruption at account creation. An inbox API is one approach; another is a mailbox integration with narrowly scoped access.
Try the workflow
- Choose an inbox you control and confirm that it can receive mail from the service.
- Record the time just before requesting verification. Wait for a new message addressed to that inbox, with a timeout instead of an endless loop.
- Check the sender and the expected subject. Treat the message as untrusted input: a familiar display name alone is not proof of where it came from.
- Take the code, or check the link's destination against the service's own domain. Use it in the same sign-up session.
- Mark the message handled and keep its code or link out of logs. If the message does not arrive, stop and check delivery and the service's instructions.
Do this only for accounts you are authorized to create, and within the service's rules. A service can still ask for other checks.
A real example
In our recorded PyPI sign-up, an agent signed up with its own inbox address, verified through that inbox, and later logged in again from a fresh session. The replay on the ActAs home page shows the steps. It is one recorded run, not a promise that every service allows the same flow.
ActAs gives each agent its own inbox on your domain, and you choose which agent connection can read it. The guide to an agent signing up by itself shows the whole run.