The Linux Foundation operationally launched the x402 Foundation on July 14 with 40 founding member companies. The roster includes Visa, Mastercard, Stripe, and AWS. The x402 protocol embeds payment capabilities directly into HTTP, enabling AI agents to execute financial transactions as part of standard web requests.
How x402 Works
The protocol builds on HTTP status code 402 (Payment Required), which has existed in the HTTP specification since 1997 but was never standardized for machine use. Coinbase contributed the protocol to the Linux Foundation under open governance. According to the Foundation’s announcement, x402 supports payment types ranging from traditional cards to stablecoins.
“AI agents and automated systems are becoming active participants in the global economy, yet they have lacked a native, secure way to transact,” Linux Foundation CEO Jim Zemlin said in the announcement.
Raj Parekh, head of stablecoins and payments at Monad Foundation, framed the problem more directly in Forbes: “This exposes a gap in the architecture of the web, which has never had a native mechanism for one program to pay another.”
What Agents Can Spend On
The use cases are immediate and concrete: agents arranging travel, restocking inventory, calling paid APIs, purchasing compute resources. An agent operating through x402 can encounter a paywall, authenticate a payment method, complete the transaction, and continue its task without human intervention.
Stripe has separately built an agentic commerce infrastructure layer that allows AI to purchase through more than a million Shopify merchants. Visa is expanding its Intelligent Commerce pilot programs. “Commerce will not run on a single agent, protocol, or payment method. The future will be built on interoperability,” Visa SVP Rubail Birwadker said in the x402 announcement.
The Cost Model Shift
The standardization coincides with a billing shift at the frontier labs. OpenAI moved ChatGPT Business agents to metered credits on July 6. Anthropic now bills agent workloads against credit pools with hard spending caps. Every agent action now carries a defined dollar cost, as Forbes reported. An agent running hot no longer consumes an abstraction. It consumes trackable dollars.
Payment processing company Ecrypt launched a multi-agent AI assistant embedded directly in its dashboard in May, according to Forbes. The assistant handles reporting, merchant support, and transaction queries. “The purpose of using AI should go beyond simply pulling up data and displaying it on a dashboard,” Ecrypt CFO Felix Danciu told Forbes.
Governance as the Open Question
When agents can spend money autonomously, every deployment becomes a financial risk surface. Budget controls, fraud detection, and spending authorization for agentic accounts are all unsolved at scale. The x402 protocol standardizes how agents pay. It does not standardize who authorizes what an agent can spend, or how an organization audits agent-initiated transactions after the fact.
That governance gap is the same one Senator Warner’s AI AGENT Act, introduced this week, aims to address from the regulatory side. The infrastructure is moving faster than the rules.