To get an idea of how big a deal AI-based commerce could be for crypto, ask entrepreneurs and developers involved in digital assets, particularly stablecoins. They’ll happily tell you blockchain-based money is the natural fit, an essential element in the mix and so forth.
Their logic is simple. Over the past few years, stablecoins — mostly digital versions of the dollar on public blockchains like Ethereum — have begun eating into the global payments industry. And while they’ve proven to be faster and cheaper than traditional bank transfers, it’s in the new world of autonomous, micro-transacting AI agents that they will shine.
That, at least, is the view of companies like Circle Internet (CRCL), the creator of the second-largest stablecoin, and technicians at crypto exchange Coinbase (COIN), which has led engineering on x402, a payments protocol designed for use by autonomous AI agents in a field becoming known as agentic finance.
Just as 24/7, frictionless, cross-border payment has been a growth area for stablecoins, agentic commerce has particular requirements that the dollar-pegged tokens meet, according to Dante Disparte, Circle’s chief strategy officer and head of global policy. Those include the ability to program the coins so they transfer only when particular conditions are met and to daisy chain, or compose, a set of actions that occur on receipt of a token.
“Firstly, you have to be able to exploit the otherwise really innocuous features of stablecoins, which is programmability and composability,” Disparate said in an interview. “Number two, where the stablecoin lives, the physical blockchain ledgers themselves, are the common reference point the agents will turn to.”
The crypto industry, however, is viewed with, if not suspicion, then at least circumspection, among some AI developers. For example, Peter Steinberger, the creator of AI agent OpenClaw, is publicly opposed to crypto, so much so that he refuses to engage in any further commentary on the subject and declined to comment for this article.
While crypto’s bullishness on AI is one end of the spectrum, consider the other side, said Sean Neville, co-founder of Catana Labs, a builder of agentic finance infrastructure that last year raised $18 million in seed funding led by a16z.
“I’ve worked with people who are more in the AI developer and engineering community that have a very low opinion of crypto,” said Neville, who is also a co-founder of Circle, in an interview. “I think stablecoins have achieved some escape velocity, but the AI developer community in particular has a negative view of crypto, because of things like memecoins and Ponzi schemes and whatnot.”
Untouched by human hands
A key feature of agentic finance is that it involves micro-transactions, or nano-payments, some of which take place between AI agents with humans somewhere in the background.
This is quite different from using Chat GTP as a front-end for a shopping cart and plugging a credit card into it, though, in the near term, agentic systems will access both crypto and cards, Neville said. Agentic payments are likely to be high-frequency transactions in the fractions-of-a cent range that credit card networks will struggle to handle.
“Over time, I do think that there are significant advantages in stablecoins and blockchain rails that are much more natural fits for agentic flows beyond just the retail commerce use case,” Neville said. “If AI is doing things like leveraging 24/7, programmable rails to stream different kinds of money around the world, across borders, it’s just difficult to do that with anything other than stablecoins.”
With clear regulatory guidance for stablecoins finally coming in the U.S., there are potentially more pressing questions for AI agents around fragmentation and conflicting protocols jockeying for position, Neville said.
“There’s a bunch of different ways for agents to pay each other, but if they can’t all agree on how payments should work, then it’s difficult to bootstrap marketplaces, whether they’re using micro payments or not,” he said. “I would love to see something like an SSL equivalent emerge for agents, and it would be great to see a standard that nobody owns, so that we could all kind of build on the same interoperable standard.”
SSL, or Secure Sockets Layer, is a standard technology that encrypts the connection between a web server and a browser.
Stablecoin-friendly option x402, which is often cited in the debate, has caused some people to get hung up on the protocol’s transaction volume from one month to another, said Erik Reppel, head of engineering for Coinbase Developer Platform and an x402 founder. He said his focus is firmly on looking ahead at a whole category of commerce that will hugely disrupt the internet’s existing advertising marketplace.
“I think the thing people haven’t quite realized is that we’re going to break the fundamental economic model of the internet, moving from browsers and you visiting the website of the person who’s publishing content, to consuming things through your agents and your chat interface,” Reppel said in an interview.
The few cents paid by an agent crawling a website, equivalent to the value of an advert flashed before a human’s eyes, could in theory be accomplished by spinning up lots of virtual cards, if a developer has a relationship with, for example, Visa, Rappel said.
“But anyone can program stablecoins,” he said. “Anyone in the world can spin up as many wallets as they want, and then just use wallets as the way to fully isolate funds for an agent. What we want is agents to have isolated, programmable funds, where your agent can’t spend into your credit card limit and can’t access your credit card.”
Catena’s Neville said the company is grappling with squaring regulated money transmission with a sea of agents and bots that have no financial identity. The goal is to keep the bad bots out, he said, while identifying and allowing the ones you want, while giving them specific guidelines and policies they can’t escape.
“The way to handle that is programmable money, because we can leverage cryptography to ensure verifiability and auditability and so on,” Neville said. “It’s effectively identity and policy controls so agents can operate within the rules, regardless of which protocol or which wallet or account infrastructure they happen to be using.”
