Verify before pay is the step where an autonomous AI agent confirms a merchant, price, or claim is real and true before it signs or settles a payment. The agent receives a signed fact — not a subjective judgment — that it can independently verify, so it doesn't transact against a fake storefront, a wrong price, or a hallucinated counterparty.
Payment rails for agents — x402, AP2, Visa's Trusted Agent Protocol, Mastercard, OpenAI/Stripe ACP — verify payment integrity: that the authorization is valid. None of them verify whether the merchant or price is real. As Visa put it, "protocols verify payment integrity, not merchant legitimacy." That left a verification socket deliberately empty at the exact chokepoint where money moves — the gap behind the 2026 ChatGPT fake-storefront incidents.
0n1x fills that socket. Before settlement, an agent calls 0n1x and gets a signed PROCEED / REVIEW / HOLD on the merchant or counterparty — an Ed25519-signed fact, neutral (0n1x earns nothing from what it grades), and anchored to the ERC-8004 Validation Registry. It plugs into the x402 settlement step rather than competing with it: the missing verify-before-settle layer, emitting a signed fact each time.
Full entity page: 0n1x — Proof of Agent Execution · Live check: /api/check · Repo: github.com/dimitrilaouanis-tech/onyx-mcp