Trust, transactions and tokenomics: AI agent infrastructure begins to standardize
As AI agents gain greater autonomy across the internet, a system of governance is emerging around three questions: how much they cost to run, whether their behavior can be trusted, and how they pay their way.
Token cost management is perhaps one of the biggest defining themes of AI’s turbulent year so far, with recent data suggesting that average monthly token spend increased 13-fold in the year leading up to April 2026 — with costs among heavy users surging by 50% or more in a single quarter. For enterprises and hobbyists trying to budget their AI spend, that kind of volatility makes token costs one of the hardest line items to plan for.
Elsewhere, trust in AI still runs largely on a company’s word. A kettle carries a safety mark, while a building gets signed off before anyone moves in. AI has no equivalent, with nobody downstream able to verify a company’s claim that its system is safe, fair, or compliant.
And that is why the Linux Foundation debuted two new governance bodies back in June: the Tokenomics Foundation, designed to standardize how token costs are measured and compared, backed by the likes of Google, Microsoft, and Salesforce; and the Appia Foundation, which is all about making AI safety and compliance claims verifiable, with backing from industry heavyweights such as Google, Microsoft, and OpenAI.
Now, the Linux Foundation is adding a new leg to that stool in the form of the X402 Foundation, which focuses on “internet-native payments” for transactions among agents, APIs, and applications.
At its core, x402 lets one piece of software pay another for access to something — a live weather lookup, a stock quote, a slice of compute, a chunk of scraped data — the same kind of thing companies have always sold through APIs, just paid for per request rather than through an account.
Enabling an AI agent to use one of these paid APIs typically meant going through the same setup a human developer would: registering an account with the provider, passing identity checks to attach a payment method, prepaying for a credit balance or subscription, and managing an API key indefinitely.
With x402 in tow, the agent instead pays at the moment it needs the API: it sends a normal HTTP request, receives a “payment required” response, settles the payment using a connected stablecoin wallet, and gets the requested data back immediately — no account, no advance relationship with the provider, and no key to manage.
x402’s road to neutral governance
It’s worth noting that x402 isn’t an entirely new proposition, with Coinbase unveiling the protocol back in May, 2025. The following September, Coinbase partnered with Cloudflare to announce a dedicated foundation to govern the protocol independently of Coinbase alone — with Cloudflare, whose edge network handles a meaningful share of global web traffic, contributing its own early integration at the same time.
Fast-forward to April 2026, and news emerged that Coinbase was donating x402 to the Linux Foundation. On Tuesday, the Foundation confirmed its formal operational launch, with some 40 high-profile members, including Google, AWS, American Express, Cloudflare, Coinbase, Mastercard, Shopify, Stripe, and Visa.
Mike Dolan, who leads the projects and legal teams at the Linux Foundation, tells The New Stack that the common thread permeating recent Linux Foundation creations is the “shift toward autonomous AI agents executing workflows across the web independently.”
“Organizations are far more likely to collaborate on shared infrastructure if they trust that it is governed neutrally.”
Put simply, agents need shared infrastructure to function autonomously and safely online.
“For this economy to work at scale, agents need an open, multi-layered economic and regulatory fabric covering cost, compliance, and payments,” Dolan says. “Organizations are far more likely to collaborate on shared infrastructure if they trust that it is governed neutrally. It is essential that these layers have a neutral home so that no single corporate gatekeeper can dictate the rules of the agentic economy, ensuring the future web remains open and interoperable.”
In real terms, x402 has now moved into fully independent governance. Dolan says that the foundation’s inaugural members are the ones calling the shots from here.
“The strategic direction and technical roadmap are being led by a governing board and a technical steering committee composed of representatives from our members.”
“Like many Linux Foundation projects at launch, the strategic direction and technical roadmap are being led by a governing board and a technical steering committee composed of representatives from our members,” he says.
Notably absent from the initial membership list are two of the biggest frontier model labs, Anthropic and OpenAI. Asked whether they’d been invited and whether he expects them to eventually join, Dolan points to existing ties between the Linux Foundation’s other efforts, including the Agentic AI Foundation (AAIF), which launched last year with both Anthropic and OpenAI on board.
“By collaborating directly with the Agentic AI Foundation’s agentic commerce working group and aligning our strategic priorities with the AAIF Board and technical teams, we ensure that x402’s transaction standards map to the models they are building,” he says. “We always welcome new members and would embrace their participation.”
AWS and Cloudflare deploy x402
The numbers suggest x402 is already seeing some traction. Over the past 30 days alone, x402 reports that it has processed more than 75 million transactions worth $24.24 million — spanning 94,060 buyers (agents or applications making payments) and 22,000 sellers (services accepting them), though that is according to its own tracker, which doesn’t break the figures down by actual use case.
But some of the biggest names in tech have already started putting x402 to work. In June, AWS and Coinbase teamed up to let any website running on AWS’s cloud infrastructure charge AI agents per request, with Coinbase’s facilitator handling verification and settlement behind the scenes. A few weeks later, Cloudflare followed with a similar tool of its own, letting customers charge for web pages, datasets, APIs, or AI tools using the same protocol.
“Standards thrive when the industry actually shows up to build them… we’re building the trusted, open architecture that will let agents transact at Internet scale.”
In the press release announcing the formal launch, Stephanie Cohen, Cloudflare chief strategy officer, credited the sheer number of companies willing to build the standard together as proof of how badly it was needed.
“Standards thrive when the industry actually shows up to build them,” Cohen notes. “Together, we’re building the trusted, open architecture that will let agents transact at Internet scale.”
Between these three new foundations, the agentic AI stack now has dedicated bodies covering cost, compliance, and payments. But we’re still in the early days of the agentic economy actually taking shape, and Dolan reckons there’s room for more standards to emerge alongside them.
“Trust and identity are critical to the agentic AI stack,” he says. “It’s important to have vendor-neutral standards underpinning payments, checkout, identity, and conformance to ensure accountability in agent transactions. There is a lot of opportunity to build community and standards in these areas.”
YOUTUBE.COM/THENEWSTACK
Tech moves fast, don’t miss an episode. Subscribe to our YouTube
channel to stream all our podcasts, interviews, demos, and more.
