Checkout.com has published a clear primer on a term that now means two different things. In What are A2A payments?, Huilin Tu uses A2A in its original sense, account-to-account: “Account-to-account (A2A) payments move funds between bank accounts through an interbank payment system.” Not agent-to-agent, which is what the same three characters mean in half the posts on this site. The piece is a map of the national schemes, FedNow, RTP and ACH in the United States, Faster Payments, Bacs, CHAPS and Variable Recurring Payments in the UK, SEPA and Wero in Europe, Pix in Brazil, PayNow in Singapore, AANI in the UAE, BLIK in Poland, and Interac in Canada, with one caution running through all of it: “A2A payments typically require a lot more trust between the sender and receiver of funds than credit or debit card payments, as transactions aren’t always easy to reverse and the sender usually cannot claim a chargeback.”
The one headline number is Wero’s, 53 million users by May 2026 after a July 2024 launch, and it is a scheme’s figure reported by a processor. The part that survives is the table, which lists every scheme’s settlement time next to its reversal rights, and it is the most useful thing in the post for anyone building against it. Most of the instant rails are irrevocable once sent. SEPA Credit Transfer has no reversal, only a recall process for errors. Bacs Direct Debit and SEPA Direct Debit carry consumer refund rights. ACH Debit allows dispute, refund, and reversal. For an agent moving money, that table is the difference between a mistake that can be undone and one that cannot, and I would like to see it published as data, not a web page.
Here the catalog has to answer for itself. The Checkout.com provider page lists 17 API pages, among them the Checkout.com Payments API, the Checkout.com Payouts API where bank payouts live, and the Checkout.com Payment Sessions & Flow API behind the Flow interface the post names. But the contracts behind the score are five stubs of about 2 KB each, 13 paths in total. Checkout.com’s API reference portal loads a 3.8 MB OpenAPI 3.0.1 document from its own host, 184 paths and 227 operations, with an idempotency key header in the contract. The catalog never found it, because the spec URL is only visible from inside the portal. That is our miss, and it is filed.
So read the measurement as a measurement of what we captured. The Kin Score is 18.7, emerging band, with contract quality at 2.1 and contract governance at 0.0, numbers that describe our 13 stub paths, not Checkout.com’s 227 operations. The Agent Readiness score is 31.1, agent-ready even on the thin record, with the MCP server, reversibility, and error semantics lit. Idempotency is unlit, and the contract Checkout.com actually serves carries the idempotency header that would light it. The honest reading of this record is that a payments company which documents reversal rights country by country has been scored on the wrong contract, and the next rescore will look very different.