Anchorage Digital is positioning bank accounts for AI agents as the missing link between conversational software and autonomous commerce, arguing that agents will need to receive, hold and send money across both cryptocurrency networks and conventional payment systems.
Nathan McCauley, chief executive of Anchorage Digital, outlined that view during the Wyoming Blockchain Symposium 2026, where he described a shift from AI tools that offer recommendations to systems that can complete transactions for customers and institutions. The company calls the model “agentic banking” and says it is building controls intended to let organizations fund AI agents while retaining authority over how those agents use money.
McCauley’s argument centers on a practical constraint: an AI agent that can find a product, negotiate a service or identify an investment opportunity still needs a recognized financial identity and access to payment infrastructure before it can execute a purchase. That requirement becomes more complicated when transactions move between stablecoins, bank payments, card networks and other systems that were built around human account holders.
“Know-your-agent” controls
Anchorage Digital has framed its proposed answer around a “know-your-agent” framework, extending familiar financial controls into a setting where software may act with delegated authority.
The concept would give organizations a way to identify an agent, set permissions, fund its account and monitor its activity. In practice, those controls could determine whether an agent can make only small recurring payments, use a limited balance, transact with approved counterparties or require human approval above a spending threshold.
Such safeguards would address a central obstacle for autonomous financial software. Traditional compliance and risk systems typically establish who owns an account and who is permitted to use it. AI agents introduce another layer: a bank or payments provider may need to determine which organization or individual authorized the agent, what instructions apply to it, and how responsibility is assigned if the agent behaves unexpectedly.
McCauley said agents will require banking capabilities beyond a payment button attached to an AI interface. Sending funds is one function, but autonomous systems would also need to accept payments, manage balances and settle transactions across different rails.
That puts custody, identity and policy enforcement at the center of the product design. A company could theoretically deploy an agent to purchase computing resources, pay a supplier invoice, collect revenue from a digital service or manage a subscription budget. Those activities require more durable financial infrastructure than a one-time wallet connection.
Anchorage’s institutional product
Anchorage Digital launched Agentic Banking in May as an institutional platform for AI-driven finance, according to company materials. The firm says the service is designed to provide trust, governance and settlement functions for clients that want to authorize agents to operate on their behalf.
The launch places the federally chartered digital asset bank in a developing contest to supply financial infrastructure for AI systems. Technology companies are building agents capable of completing more complex tasks, while payment providers, banks and blockchain developers are working on the identity, authorization and settlement layers that could allow those tasks to involve money.
Anchorage’s approach focuses on institutions rather than consumer assistants. That is a consequential distinction in the near term: businesses generally have clearer approval structures, established treasury processes and stronger incentives to limit an agent’s mandate. A corporate procurement agent, for example, can be given a fixed budget and a narrow set of approved vendors more easily than a general-purpose consumer assistant can be trusted with unrestricted spending.
McCauley referenced fictional assistants such as those depicted in The Jetsons and Marvel’s Jarvis while describing a potential consumer experience in which people delegate routine activity to software. The operational systems behind those assistants, though, would need to resolve questions about account ownership, payment finality, fraud and dispute handling before they can operate across mainstream commerce.
Stablecoins offer an early payment rail
McCauley said AI agents can already access some blockchain-based services and use stablecoins, while much of the economy remains tied to non-crypto payment networks. Anchorage Digital’s stated goal is to support agent activity across both environments rather than treating blockchain settlement as a standalone system.
That approach reflects where agent-driven commerce may first find practical use. Public blockchains can settle transactions programmatically, and stablecoins can give automated services a dollar-denominated medium of exchange without requiring every payment to pass through traditional banking hours or card authorization flows.
Network fees offer one example of the small, machine-generated payments McCauley expects agents to make. On blockchains, users pay fees to submit transactions or use network resources. An agent could apply the same model to digital services: paying for one API request, one data query, a unit of storage or a short burst of computing power instead of taking out a monthly subscription.
Micropayments have long been proposed as an alternative to bundled digital services, yet high transaction costs and fragmented payment systems have limited their use. Automated agents could create more demand for such payments because software can make frequent, low-value decisions without the friction a person faces when repeatedly authorizing purchases.
The model also creates constraints for blockchain networks seeking to serve that activity. Agents making numerous low-value payments would be sensitive to fees, transaction delays and failed settlements. They may also need reliable ways to manage stablecoin balances, select payment routes and verify counterparties without exposing the organization that deployed them to uncontrolled risks.
Banking remains the harder integration problem
The larger challenge may lie outside blockchain networks. Most salaries, merchant payments, card purchases and business invoices continue to move through bank accounts and established payment rails. An agent that operates only with on-chain assets would have limited reach unless businesses and consumers increasingly accept direct blockchain payments.
McCauley’s comments suggest Anchorage Digital sees bank-account functionality as the bridge between autonomous crypto transactions and ordinary commercial activity. If that infrastructure gains acceptance, an agent could use stablecoins for automated on-chain services while also interacting with businesses that rely on conventional payment systems.
The immediate test will be whether institutions want to give software controlled access to money before regulatory, operational and liability standards are fully settled. Anchorage’s proposed governance layer is designed for that gap: it would let organizations experiment with autonomous payments while keeping spending limits, permissions and settlement under institutional control.
See how AI transforms finance beyond banking—explore our insights in the next evolution of AI in banking.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
