Checkout.com Is Building Its Own Agent Infrastructure While Backing Three Rival Protocols

Checkout.com launched its own MCP Server on September 1, 2026, integrating payment tools into Claude Code, Cursor, and GitHub Copilot with a claimed 81% reduction in integration time, while simultaneously backing three rival agentic-commerce protocols: Mastercard Agent Pay for Machines, Google's AP2, and the Agent Commerce Protocol. The processor's multi-protocol hedge contrasts with Stripe's April 2026 full-stack SPT approach, as Checkout.com's Agentic Commerce 2026 report found 42% of merchants testing agentic workflows but only 3% of transactions involving agents, with integration costs of $5,000 to $500,000 per standard. "Agentic commerce is reshaping the checkout moment, and Google's Agent Payments Protocol (AP2) is a pivotal step forward," said Meron Colbeci, CPO at Checkout.com.

Agentic commerce is currently a study in mismatched expectations. The Agentic Commerce 2026 report https://www.checkout.com/report/agentic-commerce-2026 notes that 42% of merchants are testing agentic workflows, yet a mere 3% of transactions actually involve agents. While 70% of industry leaders call this the most disruptive force in commerce, the reality is a fragmented mess of competing protocols. This protocol proliferation tax https://forkast.news/5-checkout-standards-3-adoption-the-protocol-proliferation-tax-killing-agent-commerce/ creates a financial burden for processors, with integration costs ranging from $5,000 to $500,000 per standard. For a payment processor, the math is simple: support everything or risk obsolescence. Checkout.com is currently balancing proprietary infrastructure with broad-spectrum compatibility. On September 1, 2026, the company launched its own MCP Server https://www.checkout.com/blog/building-trusted-ai-infrastructure-mcp , designed to integrate payment tools directly into development environments like Claude Code, Cursor, and GitHub Copilot. By enabling natural-language queries for API schemas, payment status, refunds, and disputes, the company claims an 81% reduction in integration time. The goal is to establish a native payment layer for merchant agents, handling onboarding, credentials, and consent management. Building a proprietary layer is only half the strategy. Checkout.com is simultaneously positioning itself as neutral middleware by supporting three competing protocols. The company is a partner on Mastercard Agent Pay for Machines, which launched in June 2026, and is a supporting partner for Google’s AP2 https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol , announced in September 2025. Additionally, the company adopted the Agent Commerce Protocol ACP . It is worth noting that the ACP adoption dates back to November 2025 — a fact that recently resurfaced via social media, despite being ten months old. This approach contrasts sharply with the strategy employed by Stripe. At its April 2026 Sessions event, Stripe unveiled a full-stack agent commerce approach https://forkast.news/stripe-just-built-the-full-stack-agent-commerce-infrastructure-in-one-day/ , making every merchant agent-ready by default through its SPT protocol https://forkast.news/stripes-shared-payment-token-is-quietly-becoming-the-default-money-layer-for-agent-commerce/ , which covers major AI providers like OpenAI, Microsoft, Meta, and Google. While Stripe aims to own the stack, Checkout.com is betting that the market will remain multi-polar. The distinction is structural. Stripe operates as a platform; Checkout.com functions as a processor. By maintaining support for multiple protocols, Checkout.com is hedging against the possibility that no single standard will achieve total dominance. If a merchant chooses to use Google’s AP2 or Mastercard’s Agent Pay, Checkout.com remains the underlying processor, regardless of which protocol wins the interface war. This is a classic processor play, but it faces a difficult reality. With 89% of merchants preparing for agentic commerce and 72% fearing that consumer adoption will outpace their own readiness, the pressure to simplify is immense. The question remains: will merchants prefer the convenience of a single-vendor, full-stack solution like Stripe’s, or will they prioritize the flexibility of a processor that remains agnostic to the underlying agentic protocol? “Agentic commerce is reshaping the checkout moment, and Google’s Agent Payments Protocol AP2 is a pivotal step forward,” said Meron Colbeci https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol , CPO at Checkout.com. “At Checkout.com, we’re proud to support open protocols that strengthen trust and give merchants the flexibility to meet their customers where they are, however they want to shop.” Matthieu Barral https://www.mastercard.com/news/press/2026/june/mastercard-launches-agent-pay-for-machines/ , Global Head of Partnerships at Checkout.com, emphasized that the Mastercard partnership connects their “industry-leading developer and security platform with world-class payments infrastructure to power the next era of machine-to-machine commerce.” These statements underscore a commitment to being the plumbing beneath the agents, rather than the agent platform itself. Ultimately, the success of this strategy depends on whether the protocol proliferation tax https://forkast.news/5-checkout-standards-3-adoption-the-protocol-proliferation-tax-killing-agent-commerce/ remains high enough to justify the complexity of a multi-protocol hedge. If the industry consolidates around one or two standards, the value of being a neutral middleware provider may diminish. If, however, the market remains fragmented — and developments like the cross-network know your agent framework https://forkast.news/visa-mastercard-and-ant-international-build-first-cross-network-know-your-agent-framework/ suggest it will — Checkout.com’s ability to process payments across competing agentic ecosystems may prove to be a durable competitive advantage. The industry is placing its bets. The question is whether the agents will ever actually spend.