On March 16, we published The Agentic Commerce Stack Is Complete. Nobody Agreed on It., mapping every production protocol across discovery, trust, processing, and settlement. The thesis was straightforward: every layer of the stack had a production implementation, but nobody had agreed on which protocols go where, and the standards war was underway.

That thesis lasted about three weeks.

Between March 16 and April 2, the landscape shifted in ways that nobody predicted. Not incremental shifts. Structural ones. The x402 stablecoin settlement protocol went from a Coinbase project to a Linux Foundation initiative backed by the entire payments and cloud establishment. Mastercard's Agent Pay reached a third continent. Visa started buying what it had been selling. And Amazon, absent from the original map entirely, appeared in three different layers at once.

The stack is no longer a collection of competing protocols. It is consolidating. The question now is around what.

What Changed in Three Weeks

Let us be specific about what moved.

x402 joined the Linux Foundation. On April 2, Coinbase transferred the x402 protocol to the Linux Foundation as open source, launching the x402 Foundation. The founding coalition reads like someone merged the boardrooms of payments and cloud: Stripe, Visa, Mastercard, Google, AWS, Microsoft, Cloudflare, American Express, Shopify, Circle, and Polygon Labs. That is 12 founding members. Card networks, cloud providers, processors, and crypto-native companies sitting at the same table for a stablecoin settlement protocol.

Pause on that. Visa and Mastercard, whose entire business model depends on interchange revenue from card-based settlement, are founding members of a protocol designed to settle payments via stablecoins. Either they believe x402 will complement their rails, or they believe it will compete and they want a seat at the table regardless. Both interpretations are interesting.

The protocol has processed over 15 million transactions on Base, but daily volume still sits at roughly $28,000. That gap between institutional backing and actual usage is one of the most telling data points in the entire agentic commerce space.

Mastercard expanded Agent Pay to Hong Kong. After launching in Latin America (which we covered at the time), Mastercard pushed Agent Pay into Hong Kong. Three continents in one month. Latin America, North America, and Asia. That is not a pilot programme. That is a global rollout happening at payments-infrastructure speed, which is to say, far faster than anyone expected.

Visa became a Ramp customer. This one is worth reading twice. Visa signed on as a customer of Ramp, using Ramp's AI agents for its own corporate bill payments. The Trusted Agent Protocol that Visa built for merchant-facing commerce is now integrated into Ramp's platform for B2B corporate payments. The numbers are real: 85 percent invoice coding accuracy, $1 million in fraud flagged within 90 days, across Ramp's 50,000+ customers and $100 billion+ in annual purchase volume.

Why does this matter? Because TAP was designed for a world where agents buy consumer goods. Visa just demonstrated that the same protocol works for corporate expense management. The addressable market for agent authentication just expanded from retail to the entire B2B payments stack.

Amazon entered the picture. The original protocol map did not mention Amazon once. That changed when Amazon committed $50 billion to OpenAI ($35 billion contingent on IPO) alongside a $100 billion AWS cloud deal over eight years. AWS is also a founding member of the x402 Foundation. In three weeks, Amazon went from absent to holding positions in settlement (x402 via AWS), AI platform (OpenAI investment), and cloud infrastructure (the compute layer everything else runs on).

OpenAI hit 900 million weekly users. Up from 800 million when the original article published. OpenAI closed a $122 billion funding round at an $852 billion valuation. The company that controls the largest consumer AI surface, the surface where "Buy it in ChatGPT" runs, is now valued higher than most of the payment companies trying to process its transactions.

The Updated Protocol Map

Here is the updated table, with a new column tracking what changed.

| Layer | Protocol | Owner | Status (March 16) | Status (April 4) | |-------|----------|-------|-------------------|------------------| | Discovery + Intent | UCP | Google + 20 partners | Production | Production | | Discovery + Checkout | ACP | OpenAI + Stripe | Production (US) | Production (US) | | Agent Authentication | Trusted Agent Protocol | Visa + Cloudflare | Production | Production + B2B (Ramp) | | Consumer Authorisation | Verifiable Intent / Agent Pay | Mastercard | Production | Production (3 continents) | | Processing (Acquiring) | Visa Acceptance Platform | Fiserv | Deploying (Europe) | Deploying (Europe) | | Processing (Issuing) | KYA / Agentic Commerce | FIS | Q1 2026 | Shipped | | Settlement (Fiat) | Card rails | Visa / Mastercard | Production | Production | | Settlement (Stablecoin) | x402 | Coinbase + Cloudflare | Production | Linux Foundation (12+ founding members) | | Settlement (Bridge) | AP2 | Google | Production | Production | | Cloud / Infrastructure | AWS | Amazon | Not in original | x402 founding member, OpenAI cloud partner |

Ten rows now. The original had nine.

The most significant changes are not new entries. They are status changes. x402 going from a Coinbase-led project to a Linux Foundation protocol with Visa, Mastercard, and Stripe as founding members is a category shift. Visa's TAP expanding from merchant authentication to B2B corporate payments is a scope expansion. Mastercard's Agent Pay reaching three continents is a deployment velocity that most payments infrastructure takes years to achieve.

In three weeks, the stack went from "complete but uncoordinated" to "consolidating around institutional consensus." That is a different story.

What the x402 Move Really Means

The x402 transfer to the Linux Foundation deserves its own analysis because the founding member list changes the dynamics of the entire settlement layer.

When we covered x402 as a protocol, the framing was straightforward: Coinbase built a stablecoin settlement mechanism into HTTP, it was clever engineering, and it was competing with card rails for agent transactions. Card networks versus crypto. Pick a side.

That framing is now obsolete. Visa is a founding member. So is Mastercard. So is Stripe, which processes card transactions for millions of merchants. The card networks did not join x402 to kill it. They joined to shape it.

This looks a lot like what happened with open banking. The banks initially fought it. Then they realised they could not stop it. Then they joined the standards bodies and ensured the specifications worked within their existing infrastructure. Visa and Mastercard joining x402 is the same play. If stablecoin settlement is coming for agent micropayments, better to write the rules than be subject to someone else's.

Google's AP2 is now positioned as the bridge between the discovery layer (UCP) and both settlement options, fiat via card rails and stablecoin via x402. That makes AP2 one of the most strategically important protocols in the stack. Whichever settlement rail wins, AP2 routes traffic to it. Google does not need to pick a side. It needs to sit in the middle.

What Is Missing (Updated)

The original article identified three gaps: unified identity, dispute resolution, and governance. Three weeks later, all three are still open. But movement is visible.

Identity remains fragmented. Visa's TAP and Mastercard's Agent Pay still operate as separate authentication systems. No universal agent identity exists. But the x402 Foundation's governance structure, with 12+ founding members across payments and cloud, could become the forcing function for cross-protocol identity standards. Could. Has not yet.

Dispute resolution is still unsolved. We wrote about the agentic dispute crisis before the original protocol map. Nothing in the past three weeks addresses it. Mastercard's Verifiable Intent remains the closest thing to an audit trail for agent-authorised purchases. The chargeback system was designed for humans. Nobody has redesigned it for agents.

Governance is starting to crystallise. The emerging On Behalf Of (OBO) framework for agentic trust and Rep. Gottheimer's letter to Anthropic on AI safety both point in the same direction. Industry practitioners and legislators are thinking about accountability frameworks for agents that act on a consumer's behalf. None of these are production standards yet. They are proposals and policy signals. But six weeks ago, even the proposals did not exist.

Security validated the warnings. The LiteLLM supply chain breach drove that home. LiteLLM, an AI gateway present in 36 percent of cloud environments, was compromised. 500,000 machines affected. Four terabytes stolen from Mercor, a $10 billion AI startup. It was the first documented use of an AI agent in a supply chain attack. Every security concern we flagged in the original "What Is Missing" section now has a case study attached to it.

The New Power Map

Three weeks reshaped who holds leverage and where.

Google is the quiet winner. UCP for discovery, AP2 bridging both settlement rails, and a founding seat at the x402 Foundation. Google does not need any single protocol to win. It needs to sit between all of them. Sound familiar? That is the search engine model applied to commerce infrastructure.

Amazon is the surprise entrant. Not building a protocol. Not launching a framework. Just writing very large cheques. $50 billion into OpenAI. $100 billion in AWS commitments. A founding seat at x402. Amazon is doing what Amazon always does: owning the infrastructure layer and letting everyone else compete on top of it.

Visa is playing both sides of the settlement question. TAP for agent authentication on card rails. Founding member of x402 for stablecoin settlement. Customer of Ramp for its own corporate payments. Visa is not betting on which settlement rail wins. It is ensuring that agent authentication runs through Visa regardless.

OpenAI is scaling the surface. 900 million weekly users. $852 billion valuation. "Buy it in ChatGPT" live for US consumers. ACP is narrower than UCP, but OpenAI controls the screen where millions of people interact with AI agents daily. Protocols matter. Distribution matters more.

Coinbase traded control for legitimacy. Transferring x402 to the Linux Foundation means Coinbase no longer owns the protocol. But a Coinbase-owned protocol with 15 million transactions and $28,000 in daily volume was going nowhere. A Linux Foundation protocol with Visa, Mastercard, Stripe, Google, and AWS as founding members has a shot at becoming the default machine-to-machine settlement layer.

What to Watch in Q2 2026

The next 90 days will tell us whether the consolidation pattern holds or fractures.

x402 Foundation governance. The founding members include direct competitors. How Visa and Mastercard negotiate stablecoin settlement rules alongside Coinbase and Circle will define whether x402 becomes a genuine standard or a committee that produces documents.

ACP international expansion. "Buy it in ChatGPT" is US-only. OpenAI has the users globally. Stripe has the payment infrastructure globally. The question is when, not whether.

Visa TAP in B2B. The Ramp integration is a proof point. If more corporate payment platforms adopt TAP, agent authentication becomes a B2B standard, not just a retail one. That changes the revenue model entirely.

Interoperability. Still the biggest gap. An agent authenticated by Visa's TAP, checking out via OpenAI's ACP, and settling on x402 stablecoin rails still requires navigating three separate protocol specifications. The x402 Foundation might become the forum where this gets resolved. Or it might just become another standards body.

The original article ended by saying the protocols were hardening and by Q4 the rules would be difficult to rewrite. Three weeks later, the hardening is happening faster than we expected. The players consolidating around x402 suggests the industry would rather negotiate inside a shared governance structure than fight a standards war.

That is either pragmatism or it is too many cooks. Q2 will tell us which.

Sources

The agentic commerce stack took a quarter to build and three weeks to start consolidating. If the biggest card networks, cloud providers, and AI companies are all sitting at the same table for stablecoin settlement, what does that tell you about where agent payments are heading, and are you positioned for it?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.