AI rules
"Always send mail from this address to a person" or "always ignore mail whose subject contains this word", checked once per message, deterministically, in the order you set, first match wins.
Plain matching, not a second model call: free to run on every message, and the message's own text never gets a vote over how it is routed, which is what a rule interpreted by an LLM would open up.
A rule can only make the outcome stricter (escalate to a person, or ignore), never force a send. The confidence threshold and Never-auto-reply list stay the backstop they always were.
More features
Watch it answer your mail for a week
It costs nothing, it sends nothing, and at the end you will know whether you want it. That is a better argument than anything else on this page.
- No card, ever
- Your own mailbox, your own mail server
- It sends nothing until you say so