A lot of business still arrives as email and leaves as email. Protege gives a company an address that an AI assistant answers — with the tools, the permissions and the record-keeping that implies.
You delegate a mail namespace — say agents.yourcompany.com — and name an
assistant inside it. Suppliers, brokers, clients and contractors write to that address the way
they already write to a person. The assistant reads the thread, uses the tools you have granted
it, and replies.
Nobody has to install anything, create an account, or learn a new interface. That is the whole point: the counterparty keeps working the way they already work.
Businesses whose work is correspondence — the back-and-forth that surrounds an order, a claim, a booking, a renewal. The parts that never fit a portal because the other side is a different company with its own systems and its own habits.
Protege is correspondence, not campaigns. Every message we send is a reply to a message a human sent to us first.
There are no mailing lists, no acquired addresses, and no marketing or bulk sending of any kind. Each customer sends only from a domain they have proven they control, authenticated with DKIM. Bounces and complaints are recorded against the customer that caused them, and a sender generating problems is suspended automatically.
An assistant with an open inbox and real tools needs to know who is asking. Protege authorises on the cryptographically verified sending domain, so what an assistant will do for a message depends on who actually sent it — not on who the message claims to be from.
Protege is being built now and is not yet generally available. We are talking to a small number of companies whose counterparty correspondence is a real cost, and shaping the first version around them.
If that sounds like your business, we would like to hear what the correspondence actually looks like — write to hello@protege.bot.