It finds the companies, reads their sites, works out what is broken, writes the email and handles the reply. Then it stops and waits for you. Nothing has ever been sent from here without a person reading it first.
1,000 companies from the map
their site, in their words
score, and reject most
one email, evidence attached
you, every time
The honest version
Volume was never the constraint. Anyone can send five thousand emails this afternoon. The constraint is that almost none of them should have been sent, and by the time you find out, the domain is burnt.
It sends whatever it writes.
A model with a send button will confidently email the wrong person the wrong thing at scale.
It cannot tell you why.
"Personalised" usually means it found their first name. Ask what it knew and there is nothing behind it.
It learns from the wrong thing.
Optimising for replies finds people who like talking. That is not the same as people who buy.
How it works
Nothing is a black box. Every agent records what it read, what it decided and what it cost, and you can open any of them.
Pulls companies matching who you said you want. You describe the shape; it finds the list.
Reads their actual website and pulls out facts, each one marked with how sure it is.
Scores against your profile and rejects most of them, with a written reason you can argue with.
Finds the one specific thing worth mentioning, then writes two short paragraphs around it.
The queue. You read the message and the evidence behind it, then decide.
Warmed mailboxes, randomised gaps, daily caps, and an unsubscribe on every message.
Classifies what came back, drafts the answer, and stops the sequence on a no.
A notetaker joins, then coaching against your claims guide and rate card.
Not as a setting you could turn off. There is no path through this system that delivers a message a person has not read. Every gate is checked again at the last moment: your send window, your caps, the suppression list, the mailbox warmup.
Suppression is permanent
One unsubscribe stops the practice, not just the address.
Every claim is checked
A quote that is not in the source is refused, not printed.
It pauses itself
Bounces climbing? The mailbox stops before your domain does.
The agents
Not one assistant pretending to do everything. Small agents with narrow jobs, each with a page explaining how it works, what bounds it, and what it refuses to do.
What it learns from
Tell it which deals closed and it re-weights your targeting against money rather than replies. Until you have enough closed business to learn from, it says so on the screen instead of pretending.
Every adjustment is shrunk halfway back to the default and capped, because a handful of wins is not a model.
Learning from closed deals
The only evidence that says the targeting was right.
Pricing
Seats are the wrong meter. One person can point this at five thousand companies, and the work is the same either way. Every plan includes every agent.
One person, one mailbox, finding out whether this works.
or $990 a year — two months free
A team running real volume, with calls and coaching.
or $2,990 a year — two months free
Outbound as a channel you actually depend on.
or $7,990 a year — two months free
Bring your own model key and we bill you less, because the biggest cost stops being ours. Go over your companies for the month and nothing breaks — we tell you, and you decide.
Answer seven questions about what you sell, paste your website, and read what it writes. If the first ten emails are not better than yours, close the tab.