AI agents have started doing the buying.

They book travel, reorder supplies, provision cloud resources, and pay for data and compute without a person clicking anything. McKinsey estimates that agentic commerce could orchestrate $3 trillion to $5 trillion in global consumer spending by 2030, and almost all of the talk about agentic payments has centered on the agent. How it proves who it is, how it stays inside a budget, how it makes a purchase.

Far less attention has gone to the other side of the transaction, the business that has to take the money and actually get paid. That side is where agentic commerce either works for you or it doesn't. Here is what it takes to accept and settle payments from AI agents, and how the infrastructure to do it came together this year.

When the buyer is a machine, the money still has to land

For the entire history of commerce, a person had to be in the loop. Someone opened a website, typed in card details, and clicked through whatever verification screens the bank put in the way. Even a subscription needed a human to start it. That assumption is breaking. Last year's AI found information and handed it back as text. The new wave acts on a user's behalf, and one of the things it can act on is payment.

From the merchant's seat, an agentic payment is a transaction started by software acting on a customer's behalf rather than a human at a checkout page.

A user tells an agent to book a flight, a hotel, or a whole trip, and the agent finds the itinerary and pays for it in one shot. The agent authorizes and sends, and your systems verify it, accept it, and settle it. What makes this workable now is timing. Payments have become instant, global, and final at the same moment agents have learned to act.

This is still early. Most people aren't buying through agents yet, and today's volume is a rounding error next to human checkout. But the shape of these payments is already visible in the first real use cases, and it is different enough that the rails are being built ahead of the traffic.

One human request can fan out into a chain of purchases

An entrepreneur tells an agent to launch an online store, and it buys a domain, hosting, images, and checkout pages from several providers inside one budget. As that behavior scales, payments stop arriving as tidy one-off events and start running closer to a steady stream, more frequent, often smaller, and spread across more rails than a checkout page was built to handle.

What has to happen before you get paid

Before a business can settle an agent's payment, the agent has to earn the right to make it. The frameworks taking shape now handle that in a few plain steps.

The agent is verified, so everyone can trust it is acting on a real customer's authority. It is given limits on what it can spend and where, enforced automatically so it can only do what it was approved to do. It connects to providers and makes the purchase. Then the payment has to settle, reliably, on whatever rail was used.

The first steps decide whether an agent can pay. The last one decides whether you get paid. Settling reliably puts four demands on the receiving business.

• Instant settlement. Agents move at machine speed and expect funds to do the same. Anything that waits for a nightly batch breaks the flow.
• Multi-rail acceptance. Agents pay with cards, bank transfers, and stablecoins. A business needs to take all of them through one integration, not three.
• Always-on handling. Payments run in the background around the clock, with no person to retry a failure or wave through an exception.
• Fraud and chargeback protection. With no human in the loop, the business needs assurance it won't absorb the loss on a disputed or bad-faith payment.

The industry just moved to close the gap

That gap is what Mastercard set out to close in June 2026 with Agent Pay for Machines, which launched with more than 30 partners, including Coinflow, Stripe, Coinbase, and Adyen. The service extends Mastercard's earlier Agent Pay program to small, high-frequency machines generating payments in the background of commerce. It verifies agents, enforces their spending limits, and, most important for the business getting paid, guarantees settlement across cards, accounts, and stablecoins.

Coinflow joined as a launch partner because acceptance and settlement is the company's home turf.

Now we're thrilled to partner with Mastercard to bring that same simplicity to agentic commerce, enabling businesses to accept payments from AI agents as effortlessly as they do from humans.

That bridge is the real work. An agent might pay in a stablecoin while the business wants dollars in its bank account, or pay on a card rail that settles on its own clock.

Coinflow sits in the middle, taking the payment on whatever rail the agent uses and settling it to the business right away through one integration, with the conversion between stablecoin and traditional rails handled underneath. Built-in fraud and chargeback protection covers the risk, so a business can take an agent's money without picking up something new to worry about.

And Mastercard wasn't the only place that work showed up. Within days, Coinflow also partnered with Tempo, the payments-focused blockchain created by Stripe, to power the card funding layer behind the Machine Payments Protocol. That standard lets people prepay for AI compute and lets their agents spend it, paying for services on their own. Two of the biggest infrastructure bets in agentic commerce launched two weeks apart, and Coinflow is embedded in both.

Settlement is where agentic payments succeed or stall

Acceptance gets the headlines, but settlement is the harder part. A payment an agent sends in a second loses its value if the business waits two days to see the funds, and the math only gets worse as the amounts shrink.

This is why stablecoins sit at the center of nearly every agentic payment framework announced this year. They move and settle value close to real time, they are programmable enough for continuous machine-driven activity, and they keep the cost per transaction low enough that small payments still make sense. Bridged to traditional rails, they give a business instant settlement without making it hold or manage digital assets itself.

The Tempo partnership shows what that looks like in practice. A developer can fund an agent with an ordinary card, and the merchant on the other end settles in stablecoins in seconds, with finality. Coinflow is the card-to-stablecoin bridge that makes both ends work, handling card acceptance across more than 170 countries and settling the funds instantly underneath. It is the same capability Mastercard leaned on, applied to a second major standard.

Getting paid by machines, as easily as by people

Agentic commerce will reward the businesses that can take an agent's money as easily as a person's. That comes down to two things. Accepting payments across every rail, and settling them instantly with enough control to keep both sides accountable.

For years, Coinflow has let merchants accept payments instantly in whatever format a buyer chooses, which is exactly the flexibility agents will demand now that they are not bound to a card and a checkout page. That groundwork is the reason Coinflow sits inside both Mastercard's machine-payments network and Tempo's Machine Payments Protocol.

One integration brings card and bank acceptance, instant stablecoin-powered settlement, and built-in fraud and chargeback protection, everything a business needs to accept and settle payments from AI agents without rebuilding its stack.

Want to accept payments from AI agents as easily as you do from people? Talk to the Coinflow team.