Step into Northwind Labs, a live sample startup already loaded with real finances and documents. Ask its AI operator anything — then watch what happens the moment it wants to change something.
This preview runs in your browser. In real use, your own Claude connects to AMOS and operates it — the business lives in AMOS, not in the chat.
Every consequential action becomes a signed record.
Not a log you dig for later — a receipt written the instant it happens, in words you can read, with your approval on it.
A real sequence — the same one your own business would follow.
“How’s the business doing?” “Where am I losing money?” The AI reads Northwind’s real numbers and answers in plain language.
Ask it to change something and it won’t just do it. It stages the change and hands you a clear card: approve or deny.
Approve, and the action runs — and a signed proof receipt lands in a running ledger of everything the AI has ever done here.
The AI can read everything and propose anything. But anything that touches your money, your customers, or your data parks for a human decision first — and every decision is recorded.
AMOS holds itself to the identical rule. Every change to its own code passes a gate — tests, a diff bound by hash, human sign-off on sensitive paths — and you can see a real one inside the demo.