Policy adherence you can
prove across every layer.
The Arc Privacy Sector (APS) runs contracts inside enclaves, keeping their state confidential and syncing only what settles to the public Arc L1. But confidentiality is also a property of your code, not only of the platform: one Open getter, one event carrying an amount, one over-broad trust grant, and private value is back in the open — and the overhead you paid for privacy bought you nothing.
Concentric is the independent assurance layer for APS. It checks what confidential contracts actually reveal, bounds what autonomous agents are allowed to spend, and lets a transaction prove it followed a policy without disclosing the amount.
Leak checks for everything bound for the private chain
APS runs your contract in an enclave: storage is confidential by default, execution proceeds alongside the public chain, and only settlement is periodically synced back to Arc L1. That model gives you privacy — and a new failure mode. Every return value, event, bridge call, and trust grant is a place a confidential value can cross back into the open. Get one wrong and the contract leaks, quietly, in production.
Concentric reads the Solidity source and its @aps access policy and traces every one of those crossings before you deploy. It works statically in your browser or local environment; no contract or key leaves your device.
The split is deliberate and complementary: you use a public-chain simulator for what settles publicly, and Concentric for what stays private. And there is no point paying the overhead of privacy if the contract leaks anyway. Concentric is what makes the privacy investment worthwhile.
Give an agent a budget, not your wallet
The next thing to hold USDC on APS will not always be a person. Software agents are beginning to pay for data, compute, and services on their own via the x402 pattern that turns an HTTP 402 into a real payment. That moves the risk: the question is no longer only whether the contract is correct, but whether the agent can be made to do something it shouldn't. Recent x402 security research catalogues exactly how an agent can be coaxed past its budget, pointed at an attacker's address, replayed, or prompt-injected into draining its balance.
Concentric extends its formal core from contracts to agents. An agent's spend authority is expressed as an attenuable, verifiable capability — a bounded credential in the object-capability and macaroon lineage — and every payment is checked against it before it is signed.
The guarantees are concrete: the agent cannot exceed its cap, cannot pay an address off its allowlist, cannot replay a spent authorization, and cannot be talked out of any of it by a prompt injection. This hardens the on-chain spend-limit and allowance mechanisms emerging in the ERC-4337 / ERC-8004 agent-wallet stack, and composes with APS so a payment can be confidential as well as bounded.
An agent developer ships autonomy without handing over the keys; an enterprise puts a provable ceiling on what its fleet can move; and the ecosystem gets the assurance story that makes agentic USDC safe to turn on at scale.
Prove rules were followed without revealing the amount
APS gives you selective disclosure, but the primitive is blunt: prove you are authorized to look, and you see everything — the full amount included. For a regulated institution that is often the wrong trade-off. The auditor rarely needs the number; they need to know a rule held.
Verifiable compliance replaces show everything with prove the property. A named policy is written in a small verified policy language and compiled to either an in-APS predicate check or a zero-knowledge proof.
Confidential and provable compliance on which others can build: signed, exportable attestations for the auditor, live monitoring of trust-domain changes on deployed contracts, and ways for regulated USDC flows to live on APS without disabling confidentiality to satisfy oversight.
Privacy infrastructure: as trustworthy as the contracts it hosts
Concentric is an independent, open assurance layer that keeps that promise honest by catching leaks before they ship today, bounding what autonomous agents can spend next, and letting confidential transactions prove compliance without surrendering privacy. Builders ship with confidence, auditors and compliance teams get real evidence instead of promises, and the Arc Privacy Sector becomes a place serious USDC transactions — human and agentic — can actually settle.