How the inbox watcher works
When someone replies to your outreach, it gets caught, cleaned, and filed in your tracking sheet automatically. No checking inboxes, no copy-paste, no reply slipping through the cracks while you were busy.
A look under the hood: the watcher that never sleeps, the cleanup that throws away the noise, and the strict rule that it only ever reads your mail.
The short version
It's two parts working together. A watcher checks your outreach inboxes once a day, finds every genuine reply, strips each one down to the words the person actually wrote, and hands them off. Then an agent reads each cleaned reply, works out what it means, and writes it as a row in your tracking sheet. You wake up to a sheet that's already current, with every reply logged and sorted, and you never touched your inbox to make it happen.
That's the whole job. The rest of this page is what each part does, and why it's safe to point at your email.
The green part is the watcher, the tireless half that catches and cleans. It hands a tidy reply to the agent, which makes the judgment call and writes the row. Two simple jobs, cleanly separated.
The watcher: it catches every reply, so you don't have to
Once a day, the watcher quietly checks your outreach inboxes and pulls in everything new since it last looked. It's looking for genuine replies from real people, and it casts a wide net on purpose, so nothing real slips past.
It connects to your inboxes and reads only what's arrived since its last run, so it never re-reads the same mail twice. It keeps the obvious noise out, the automated bounce-backs and system notifications, but it deliberately doesn't try to be clever about what counts as a "real" reply. Anything that looks like a genuine response to a thread is passed through, because the cost of being too strict here is the worst possible outcome: a real reply silently dropped. Better to pass a few borderline ones along and let the agent, which actually reads the message, make the final call.
If it ever fails to run, it says so, with an alert, rather than failing quietly. A watcher you can't trust to tell you when it's broken isn't worth having.
The cleanup: just the words, none of the noise
A raw email reply is mostly clutter: the entire previous conversation quoted underneath, a signature block, a legal disclaimer, formatting junk. Before anything gets filed, all of that is stripped away, leaving only the new words the person actually wrote.
This is the quietly clever part. When someone replies, their email usually carries the whole thread quoted below their message, plus their signature, plus whatever their company appends to every email. If you logged that raw, your sheet would be unreadable. So the watcher peels it back: it drops the quoted history, removes the signature and the footer, and converts any fancy formatting down to plain text. What's left is just the reply itself, the part you actually care about, ready to be read and understood.
The same trick a person does by hand, reading past the quoted thread to find the actual reply, done automatically on every message before it's logged.
The agent: it reads each reply and files it
Once the replies are caught and cleaned, a separate agent takes over. It reads each one, decides what kind of response it is, finds the right row in your tracking sheet, and fills it in. The watcher gathers; the agent judges.
Splitting the work this way is deliberate. The watcher's job is simple and mechanical: catch everything, clean it, hand it over. The thinking, what does this reply actually mean, which campaign and which prospect does it belong to, how should it be logged, is left to the agent, which reads the message the way a person would. The agent works only from the cleaned replies it's handed; it never touches your inbox itself. So the part that reads your email does as little as possible, and the part that makes judgments never sees your raw mail at all.
Why it's safe to point at your email
The watcher only ever reads. It opens your inbox in read-only mode, never marks, moves, deletes, or labels a single message, and it has no ability to send email at all. It's a one-way mirror onto your replies, nothing more.
This matters, because handing any automation access to your email is a real decision. So the access is deliberately minimal. It reads new messages and nothing else. It cannot reply, forward, archive, or alter your mailbox in any way, there's simply no machinery in it to do so. And it keeps its own private note of how far it's read, so it picks up exactly where it left off and never processes the same reply twice, even if it's restarted. The result is an automation you can point at a live inbox without worrying it'll touch something it shouldn't.
What you get, and what it never does
You get a tracking sheet that stays current on its own: every reply caught, cleaned, understood, and logged in the right place, day after day, without you watching an inbox. What you don't get is anything sent, changed, or deleted in your name.
The whole point is to close the gap between "a reply landed" and "it's logged where it should be", the gap where real responses get missed because someone was busy. It reads, it understands, it files. It never writes back to the people in your inbox, never makes a decision about what to say, never alters your mail. The machine does the tireless watching and the tidy filing. You keep every reply, and the final word on what to do about it.