The Linux Foundation today revealed it is setting up an x402 Foundation that will oversee the development of a payment protocol for artificial intelligence (AI) agents that can be embedded into web applications.
The x402 Foundation will include Coinbase, Cloudflare, and Stripe on its governing body. Initial indications of support have also been made by Adyen, Amazon Web Services (AWS), American Express, Ampersend.ai, Base, Circle, Fiserv Merchant Solutions, Google, KakaoPay, Mastercard, Merit Systems, Microsoft, Polygon Labs, PPRO, Shopify, Sierra, Solana Foundation, thirdweb, and Visa.
Announced at the MCP Dev Summit North America conference, Jim Zemlin, CEO of the Linux Foundation, told conference attendees the goal is to provide a neutral home for advancing an x402 protocol to pay for accessing application programming interfaces (APIs) and web content using cryptocurrency such as stablecoins.
Based on an extension to the HTTP protocol originally developed by Coinbase, the x402 protocol was created to enable instant micropayments using an immutable blockchain backend that eliminates the need for signups that require someone to divulge their identity. That protocol is now being extended to support e-commerce applications involving AI agents.
It’s not clear to what degree cryptocurrencies will be used to pay to access APIs and content in the age of AI, but more organizations than ever are trying to find ways to monetize data at a time when AI agents and applications have already consumed much of the publicly available data. The x402 protocol, in theory, would provide a means to subsidize the cost of accessing private data that could be more selectively shared with, for example, an AI agent or model.
However, there are now multiple payments protocols now being advanced and there may come a day when some of the functionality embedded within them becomes more widely shared across all of them versus requiring IT organizations to deploy and maintain applications based on a divergent set of payment protocols.
Regardless of payment method, there is no doubt at this point that AI agents will be able to make purchases on behalf of individuals and organizations. Exactly how negotiations might be conducted is another matter. In theory, an organization might set up an AI agent designed to sell a product or service at the highest margin possible. A buyer, however, might have an AI agent that is instructed to acquire a good or service at the lowest cost possible. Those two AI agents at some point may need to conduct a negotiation to establish some type of discount level that works for both parties.
In the meantime, organizations should assume that any AI application involving purchasing is going to present a rich target to cybercriminals. AI agents have already shown they are susceptible to malicious prompt injection attacks that theoretically could also be used to instruct an AI agent to make an unapproved payment that once made could not be rolled back.
Hopefully, more security will be embedded within payment protocols to prevent them from being abused, but organizations should assume those efforts are for now still a work in progress.
