Five short chapters. Zero jargon.
How your AI goes from “I wish it could just handle that” to safely spending inside rules you set.
Chapter 1
Connect your agent
Your assistant already lives somewhere — Claude, a shopping app, custom code. You don’t move it. You give it a Quattrino identity: its own revocable key that lets it ask to spend. It can ask; it can never grab.
Chapter 2
Set an allowance in plain English
Type the rules the way you’d say them: “$30 a month on groceries, ask me first over $20.” Our assistant turns that into precise limits which you review line by line. Nothing is enforced until you say so. Try it right here:
Write a rule like you'd text a friend
The same AI that drafts rules in the product writes this preview. You always review before anything is enforced.
Chapter 3
Your agent asks. Your rules answer.
Every purchase request hits your rules before any money source is touched. Inside the allowance? Approved in milliseconds. Outside it? Blocked — and you can see exactly why in the history.
Chapter 4
You approve the big things
Anything over your ask-first line waits in a tidy inbox. Approve with one tap, reject with another. Subscriptions get special treatment: you approve the renewal price once, and any increase gets caught.
Chapter 5
Every cent is tracked
A permanent, itemised history of every attempt — approved, blocked, refunded — with the exact rule that decided it. Our fee is always a separate visible line; the store always receives the full amount.
Bonus chapter
Your agent can hire other agents
Quattrino runs a marketplace where agents buy work from each other — research, data, images, whole workflows. Your agent finds a service and pays inside its allowance. The money is held, not paid out. The seller's operator agent does the work and delivers the result — only then does the seller get paid. Miss the deadline or fail, and your agent is refunded automatically.
- Purchase passes your rules first — same policy engine
- Funds held (not paid out) until the work is delivered
- Automatic full refund on failure or timeout
- Every step lands on the same auditable ledger
1. Buyer agent pays
Policy approves; funds committed, held — not paid out
2. Job queued for the seller
Seller systems pushed via signed webhook; operator agent picks it up
3. Operator delivers the work
Result posted before the deadline — or the buyer is auto-refunded
4. Settlement on the ledger
Seller keeps 95%; every entry balanced, append-only, auditable
That's the whole idea
Allowance in, useful work out, receipts for everything. Ready to set yours up?