That is the road Sean Neville, co-founder of Circle and a principal architect of the USDC stablecoin, has set out to travel again with Catena Labs, a Boston-based startup that emerged from stealth May 20 with $18 million in seed funding led by a16z crypto. Catena’s goal is to build what Neville called the world’s first “AI-native” financial institution. It will be regulated, licensed and purpose-built for a future in which autonomous software agents, not humans, initiate and settle most transactions.
“The only actors that will be executing financial transactions directly will be AI actors,” Neville told Karen Webster in an interview. “We humans and businesses will be interacting with those AIs — whether you want to call them agents or agentic workflows — just as naturally as we tap a screen today.”
An economy running at machine speed demands money that can keep up, he said.
“The banking system today is designed around a brick-and-mortar assumption,” he said. “If machines are transacting at machine speed, they need money that moves at machine speed, not the rails designed for human speed. Moving money still feels more like carrying physical goods across a border than delivering information, and when AI actors are negotiating contracts, performing FX on the fly and paying for one another’s content in micro-transactions, we’ll need near-instant settlement at extremely low fees. The current system is not situated to provide that.”
Infrastructure First; Bank Next
Catena plans to tackle that gap from two directions. On the technology side, it released the open-source Agent Commerce Kit (ACK), a collection of protocols for identity, discovery and payments among AI agents. On the institutional side, the company is pursuing money-transmitter and other state and federal licenses that will let the platform settle transactions in regulated stablecoins and, eventually, hold deposits.
“Today an agent can’t be fingerprinted, so it can’t get a money-transmitter license,” Neville said. “Our answer in the near term is that we are acquiring the licenses and developing an identity protocol that allows AI actors to prove they are acting on our behalf.”
The seed round will fund the build-out of that dual track: Catena’s regulated entity and the middleware that lets agents authenticate and pay one another programmatically. The startup’s first customers will be the developers and enterprises deploying the agents themselves; ultimately, Catena wants the agents to make up its primary user base.
However, regulation looms over the model from two angles. One is policy around dollar-backed stablecoins, which Neville said he expects to clarify once pending federal legislation becomes law.
“It’s not a silver bullet for agentic commerce, but regulated stablecoins are a very important component,” he said.
The second is the absence of a framework for AI entities that, unlike people, can’t submit to know your customer checks. Catena’s proposed remedy is a cryptographic credential that ties every agent transaction to a licensed legal entity without exposing personal data. That would give supervisors an auditable trail while allowing agents to trade at scale, Neville said.
“You have to have the trust, right?” Webster said during the conversation. “You can have the technology, but you need the trust.”
Her observation underlined the largest unknown in Catena’s plan, which is whether companies, regulators and consumers will accept autonomous agents as safe stewards of money.
Banking on Trust
Catena is spending its first 18 months on data science techniques that let agents “judge one another’s output” and operate under graduated guardrails, much like a new employee earns responsibility over a 30-, 60- or 90-day review cycle, Neville said.
“The word you used — trust — is the biggest hurdle,” he told Webster. “It’s not really tech; it’s mostly trust.”
The company believes that solving agent identity and authentication will unlock new risk models that selectively block malicious bots while letting compliant ones pass, reversing the reflexive “block all bots” stance common in eCommerce fraud systems today, he said. If it works, merchants, logistics platforms and other enterprises could allow agents to chat, negotiate and settle bills without human intervention, so long as every message carries Catena’s cryptographic provenance.
Catena’s near-term critical path splits in two. The first branch is product: shipping an identity layer, integrating existing payment networks and publishing additional ACK modules “in the coming weeks,” Neville said. The second is legal: assembling the license portfolio needed to offer insured accounts, cross-border remittances and other banking functions.
Over time, Catena expects standards for agent discovery and inter-agent communication to coalesce, just as HTTP unified the early web, he said.
“Imagine if you and I wanted to build an Amazon.com before HTTPS existed,” Neville said. “HTTPS doesn’t really exist for agents yet, and that protocol layer is in the critical path.”
Investors are betting the team that helped bring mainstream finance to blockchains can repeat the feat for autonomous software. Whether Catena’s blueprint becomes the HTTP of agentic finance — or stalls in the trust gap — will depend on how quickly regulators, enterprises and end users join the journey.
For now, Neville is lacing up for the long march, including product launches to make the concept tangible, licenses to make it legal and, above all, the slow accumulation of confidence that convinces the market to let the machines take the wheel.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
