t54 and Hecto Financial Want AI Agents to Prove They Have Permission to Pay.

An AI shopping assistant might find the right product at the right price. Before it pays, the payment system still needs to establish who sent it, what it is allowed to buy and whether the purchase matches the customer’s instructions.

That is the problem t54 Labs and South Korean payments company Hecto Financial want to address. The companies announced a memorandum of understanding on October 27 to explore identity verification and transaction risk controls for AI agent payments.

The initial focus is Hecto Financial’s AI Payment Platform, which is under development, and t54’s Trustline risk management platform. Their first step will be examining whether Know Your Agent, or KYA, verification can be applied to agents using Hecto’s infrastructure.

A Spending Limit Is Only Part of the Job

Imagine asking an assistant to buy a replacement printer cartridge for less than $80. A purchase could fall within that budget and still be wrong: the cartridge might be incompatible, the seller might be questionable, or the agent might sign up for a recurring subscription you never requested.

That hypothetical example illustrates why agent payments need several checks working together. Identity establishes which agent is acting. Delegated authority establishes what the user has permitted. Transaction context helps determine whether the proposed purchase fits that permission.

According to t54, Trustline evaluates those factors before a transaction executes. The proposed collaboration would explore how those checks could fit into Hecto’s payment and settlement systems.

For merchants, the potential benefit is a clearer basis for accepting an order placed by software. For customers, it is a way to give an assistant useful purchasing authority while keeping control over how that authority is used.

Bringing Agent Controls Into Existing Payments

Hecto Financial provides payment processing, cross-border settlement, remittance and foreign-exchange services. Its existing business gives the discussions a practical focus: connecting agent verification with infrastructure that businesses already use to receive and move money.

The company has previously announced an AI payments proof of concept based on AP2 and participation in the x402 payment ecosystem, according to the announcement. The companies will also explore potential uses involving stablecoins and payment workflows connecting merchants with AI agents.

Hecto CEO Jong-won Choi said the collaboration would examine identity verification and transaction risk management suited to the company’s payment infrastructure, as agents become involved in more of the shopping journey.

The key question is how to carry the customer’s instructions through that journey—from discovering a product to requesting and executing payment.

“Agent payments require more than a payment method; they also require a process for verifying that the authority granted to an agent aligns with the actual transaction.”

— Chandler Fang, co-founder of t54

A Growing Base of Agent Activity

t54 already has activity on which to build. Its website reports more than 20 million transactions screened and over 41,000 agents verified through KYA. Those are company-reported figures for its existing infrastructure, separate from the proposed Hecto collaboration.

Blockster previously covered t54’s milestone of 10 million agent-initiated transactions on the XRP Ledger, including how software can pay for data, computing power and other services within defined spending limits.

The company is also exploring more complex financial decisions. Its planned collaboration with RiskX would test personalized investment workflows alongside checks on agent identity, authority and customer objectives.

The Hecto discussions bring that same question into commerce: how does a payment provider establish that an agent’s action is appropriate for the person it represents?

What Comes Next

The MOU sets out an exploratory collaboration. The companies will use working-level discussions to define initial use cases, technical integration points and the scope of validation. Decisions about beta participation and integration methods remain ahead.

The starting point is agent identity. Discussions are expected to expand into whether requested transactions match users’ intentions and stay within authorized limits.

For AI commerce to become useful at scale, customers need to be comfortable letting software complete purchases on their behalf. Making permissions understandable and enforceable could help turn that comfort into everyday use.