Slow at the counter
Most chains confirm block by block, so "final" means waiting minutes. Checkout and software agents cannot wait.
Proof of Balance
zkLoop Labs builds payment infrastructure for people, businesses, and AI agents. Every payment settles in about a second, carries no gas fee, and keeps balances and customers out of public view.
The problem
Card networks are fast but rigid. Public blockchains are open but slow and transparent. Businesses need both speed and privacy.
Most chains confirm block by block, so "final" means waiting minutes. Checkout and software agents cannot wait.
When every payment joins one global line, a sale in one shop queues behind a sale in another.
Shared ledgers expose revenue, prices, and customers to anyone, including competitors.
How it works
PoB replaces public numbers with mathematical proof. The network can verify every payment while learning almost nothing about it.
Balances are stored as encrypted commitments. Each payment carries a zero-knowledge proof that value is conserved and nobody spends money they do not have.
A payment is final once a quorum of independent validators certifies it, typically two network round trips. No confirmations, no reversals.
The ledger is a graph of balances. Payments touching different balances proceed in parallel, so capacity grows with hardware instead of hitting a block-size cap.
Nobody buys or budgets a gas token. Abuse is stopped by a private minimum-balance check, and fees are simple merchant settings.
A standard EVM chain for developers, a plain REST API for merchants, and an agent interface for AI assistants. Chains can run private.
The execution engine is treated as untrusted. Validators re-run and certify every transaction, so even a compromised engine cannot forge history or create money.
Programmable payments
Fuel, parking, rentals, bookings, and metered services do not know the final price up front. PoB builds preauthorization and capture into the payment itself.
The customer places funds in escrow with a merchant before the final amount is known.
The merchant settles the actual amount. The difference returns to the customer in the same step, along with any fees and rewards.
If the order is cancelled, the full hold goes back to the customer by rule, not by dispute or chargeback.
Payments for AI agents
Agents transact fast and unattended. They need answers they can act on, and users need guarantees that hold even when an agent reasons badly.
You approve an assistant once: this merchant, up to $200, for the next 30 days.
The assistant searches, gets a quote, and reserves the room. The merchant places a hold.
After check-out the merchant settles the real amount and the rest returns to you. If plans change, the hold is released.
If the assistant is ever tricked into asking for more, its token simply refuses.
Compliance
A payment system must do more than prove money was not created from nothing. The operator also needs to know who is participating and to decide which payments are allowed.
The operator verifies each customer or merchant first. Every identifier is tied to that verified identity, so generating a new key grants no new rights.
A restriction applies to everything associated with the customer. Validators stop existing value from being relabelled to evade it, without taking the customer's keys.
Validators check the eligibility of both sender and recipient before certifying a payment. Going through a different API or network route cannot bypass the check.
Users keep their wallet keys. Revocation works by refusing future certification, and services can act only through operations the user delegated.
A processing service can be allowed to combine or split balances within one identity, while a transfer to another identity needs separate approval.
Business privacy protects data from the public and other participants. The operator keeps the records its authorized services and compliance duties require.
An architecture, not an approval. This describes how PoB can support regulated operation. Adopting it is not regulatory approval, and accurate onboarding remains essential. Section 8 of the whitepaper covers the design in detail.
Where PoB fits best
An operator issues stored value that customers spend at a defined set of merchants. Malls, fuel networks, transit, campuses, games, and marketplaces all work this way, and card rails serve them poorly.
The operator sets admission and payment rules. Customers keep their own wallet keys, and the validator quorum enforces integrity.
No interchange percentage and no gas auction. A 40-cent locker rental costs what the infrastructure costs.
Merchants in the same loop cannot see each other's volumes, prices, or customers. Each reads only its own activity.
Expiry, merchant scoping, promotions, and family or fleet spending limits are rules in the contract, not after-the-fact monitoring.
Every unit is backed by an attested cash-in, and the conservation proof lets anyone audit the total without opening a single balance.
Each business can run its own private chain while one customer balance pays across all of them, with no bridges.
What PoB is, and is not
An operator runs the stack, sets admission policy, arranges backing and redemption, and chooses its validators. Users keep their keys inside that defined economy.
PoB is not an anonymity coin. It hides balances, amounts, counterparties, and activity patterns from the public and from other participants. It does not hide participants from the operator or from lawful audit.
It is not a global currency. Tokens in a loop represent attested, redeemable stored value. The system is in active development, with demonstration deployments today.
The whitepaper explains the technology end to end in plain language. It is available on request.