World Liberty Financial (WLFI), a decentralized finance company with ties to US President Donald Trump, announced the AgentPay SDK on March 19th. It is an open source software development kit designed to enable artificial intelligence (AI) agents to autonomously make payments using WLFI’s proprietary stablecoin, USD1.
According to WLFI, the AgentPay SDK promises to provide agents with the following features: Pay for services, hire resources, and resolve obligations. within the rules defined by human operators.
According to the WLFI statements, the SDK behaves as follows: Four layers working together:
Command line interface (CLI)
it is about agent pay, SDK operation tools This allows you to interact with the system directly from your terminal.
It is through this environment that payments are made, balances are referenced, policy engine rules are set, authorization requests are managed, and it serves as the link between agent instructions, condition validation, and local signatures before sending transactions to the network.
It is aimed at developers and operators who build or monitor AI agents, giving them explicit control over how and when these systems can move funds within defined limits.
Local storage
The SDK includes a component called . vault demon They describe WLFI as signing transactions directly on the user’s device. Communication between the agent and its components is done using Unix domain sockets, a technology that allows two programs to run on the same machine. exchange data without using the internetTherefore, the private key never leaves the device or reaches the WLFI server.
statement It does not specify where these keys are physically stored within the device.Whether encrypted files on disk, in memory, or on dedicated hardware components.
Operational policy engine
Before the agent makes the payment, The system evaluates whether the transaction complies with the rules defined by the operator.: per transaction limits, daily spending limits, active network.
If a payment exceeds these thresholds, the SDK does not reject the payment, but pauses it and generates a manual authorization request to the user. According to WLFI, this differentiates between routine trades that are executed automatically and risky trades that wait for a human decision.
Integrated package (skill pack)
These packages are configuration files that the SDK automatically installs into the development tools you use, such as Claude Code, Codex, Cursor, Windsurf, Cline, and Goose. According to the statement, if an agent can read the configuration file, it can use the SDK to make payments.
Role of USD1
The AgentPay SDK is built around USD1, WLFI’s stablecoin indexed to the US dollar. WLFI presents USD1 as an “operational economic layer for autonomous AI systems.” It is a unit of account designed to allow machines to pay for services, application programming interfaces (APIs), or products without the friction of traditional banking systems.
At the time of SDK release, USD1 is now available on Ethereum and BNB chains No additional configuration is required. The contract address and network parameters are already included by default.
According to the statement, BNB Chain was chosen because of its low transaction fees. As of this writing, the average cost of the BNB network isIt is about 0.07 USDfor Ethereum it is 0.01 USD.
Current limitations
The SDK has three related limitations that the statements themselves are partially aware of.
- First gas addiction. To perform a transaction, a broker must have the network’s native token (BNB on the BNB chain or ETH on Ethereum) to pay network fees. This means that operators need to manage two different assets: 1 USD for payments and a native token for gasoline. WLFI notes that it is working on implementing EIP-3009, a standard that enables gasless transactions, but has not disclosed its implementation date.
- The second is the learning curve.. Although the release describes installation as a “single command line,” working with the SDK in production requires managing local daemons, configuring spending policies, and managing wallets. This level of complexity currently makes it a tool for experienced developers rather than end users.
- The third is the risk of agent manipulation.. Although the SDK does not specify this, if the AI agent is tricked using a technique called instruction injection (Immediate injectionif a malicious attacker injects hidden commands into the content processed by the agent), he or she could attempt to execute payments until the limit allowed by the policy engine is exhausted. The statement does not address that scenario or describe any specific mitigation mechanisms.
Where WLFI points to
In addition to the launch, WLFI has outlined a roadmap that includes the implementation of EIP-3009 to eliminate gas dependence, an open standards proposal (EIP) for all AI agents to operate under the same payment scheme, and future integration with DeFi protocols to allow agents to invest or manage capital rather than just spend it. None of these developments have been dated.
WLFI is launching the first layer of infrastructure, not a complete ecosystem. Its real value will depend on how many developers adopt it, whether USD1 can establish itself as the unit of payment for AI agents, and whether current limitations, especially gas dependence and operational risk, are resolved before more established competitors occupy the space.
The release of the AgentPay SDK is part of a broader trend. Visa introduced Visa CLI, a tool for AI agents to autonomously make fiat payments using tokenized credentials on traditional rails without human intervention.
In parallel, ecosystem actors such as World (formerly Worldcoin) and exchanges such as Kraken, OKX, and BingX are developing infrastructure for agents; Focus on identity, payments and autonomous transactionsThis marks a convergence towards a system where machines operate directly in the digital economy.
(Tag Translate) Cryptocurrency
