World Liberty Financial, the Trump-backed DeFi project, has launched AgentPay SDK, an open-source toolkit that gives AI agents the ability to hold funds, execute payments, and move money across EVM-compatible blockchains using the project's USD1 stablecoin. The SDK, now live at agentpay.worldlibertyfinancial.com, includes built-in policy enforcement, human approval workflows, and a secure signing boundary that keeps private keys on the developer's machine.
The release follows co-founder Zak Folkman's March 12 tease that "some very exciting updates coming that will completely change how people think about AI agents making autonomous payments." AgentPay is WLF's answer to a question the entire stablecoin industry is now asking: who builds the payment rail that autonomous software agents run on?
What AgentPay Actually Does
The SDK is designed around a specific problem. As AI agents move from chatbots to autonomous actors that book flights, pay invoices, and manage subscriptions, they need a way to hold and spend money without handing full control to any single party. AgentPay's architecture addresses this with three layers.
First, a unified policy engine lets developers define custom rules that execute before any transaction. Spend limits, destination allowlists, and transaction thresholds are enforced automatically. An agent tasked with paying a $50 SaaS invoice cannot drain the wallet for $50,000 because the policy layer blocks it before the transaction reaches the blockchain.
Second, human approval gates trigger when actions exceed configurable thresholds. A developer building an AI purchasing agent can set a $500 ceiling for autonomous spending and require manual sign-off for anything above that amount.
Third, the signing boundary keeps private keys local. Keys are never shared with agents or plugins. The SDK uses Secure Enclave and TEE (Trusted Execution Environment) technologies for local signing, meaning the AI agent can request a payment but cannot access the underlying wallet.
The codebase is open source on GitHub, with a plugin architecture that supports custom commands, adapters, and integrations.
A Five-Way Race for Agent Payment Rails
WLF is not building in a vacuum. The agentic payments space has attracted some of the largest names in crypto and fintech over the past six months.
Coinbase shipped AgentKit, its own developer toolkit for giving AI agents crypto wallets and on-chain interaction capabilities. AgentKit supports all EVM and Solana networks and already integrates fee-free stablecoin payments. Stripe is building Tempo, an entirely new blockchain designed for programmable payments. Circle has been developing nanopayment capabilities for USDC, targeting the micro-transaction layer where AI agents might pay fractions of a cent per API call. Shopify has integrated stablecoin payments into its merchant platform.
The distinction WLF is drawing is around USD1 specifically. The fifth-largest stablecoin by market capitalization, USD1 had crossed $4.7 billion in circulating supply as of late February 2026, though Binance held approximately 87% of that supply according to Forbes reporting at the time. Whether developers choose to build on USD1 rather than USDC or USDT depends on distribution, integrations, and whether WLF's political associations are an asset or a liability in enterprise adoption.
The Policy Question No SDK Can Answer
Every agent payment SDK, regardless of issuer, faces the same regulatory uncertainty. If an AI agent autonomously executes a payment to a sanctioned entity, who is liable: the developer, the agent operator, the SDK provider, or the stablecoin issuer? AgentPay's documentation explicitly states it is provided "as is" with no warranties, and developers retain full responsibility for compliance.
This is not a hypothetical concern. The SEC and CFTC's recent joint classification of 17 crypto assets as digital commodities established clearer boundaries for tokens, but agent-initiated transactions exist in a gray area that neither agency has addressed. The stablecoin yield debate in Congress also has implications: if USD1 or any stablecoin used by agents generates yield, the compliance burden multiplies.
WLF itself has faced political scrutiny. In February, 41 House Democrats pressed Treasury Secretary Bessent about WLF's OCC bank charter application and a $500 million UAE investment. Senate Democrats have called for CFIUS review of the deal. An open-source SDK does not insulate WLF from these questions, but it does create distance between the protocol layer and the political entity.
Market Context
As of March 20, 2026, the broader stablecoin market sits near $315 billion in total supply. BTC trades at $70,228 (-1.2% over 24 hours), ETH at $2,147 (-2.1%), and the Fear and Greed Index reads 31, firmly in "Fear" territory. The market environment is cautious, which makes infrastructure launches like AgentPay less about immediate trading impact and more about positioning for the next cycle's killer use case.
The bet WLF is making is straightforward: if autonomous AI agents become a meaningful source of on-chain transaction volume, the stablecoin they default to will capture enormous network effects. Whether that stablecoin is USD1, USDC, or something built natively on Stripe's Tempo chain is the $315 billion question.
Overview
World Liberty Financial has released AgentPay SDK, an open-source toolkit enabling AI agents to hold, move, and transact with USD1 across EVM chains. The SDK includes policy enforcement, human approval workflows, and local key signing. It enters a competitive field against Coinbase's AgentKit, Circle's nanopayment infrastructure, and Stripe's Tempo blockchain. Regulatory questions around agent-initiated payments remain unresolved, and WLF's political associations add a layer of scrutiny that competitors do not face.
