Ant International, Visa, and Mastercard introduced the agentic commerce framework with a specific pitch: register an agent once, and it can transact everywhere. The idea was a single agent identity layer that would let autonomous software operate across merchants without repeated verification. The framework’s merchant-side story has a hole in it.
The framework addresses who the agent is, who authorized it, and what it can do. What it does not provide is a unified merchant-side integration standard, shared SDK, or certification path. Three frameworks are now competing for the merchant verification layer – each with different integration mechanics. Visa’s Trusted Agent Protocol relies on HTTP-layer signature verification: merchants retrieve public keys from a Visa directory and verify signatures on inbound agent requests, with sample code available on GitHub. Mastercard and Google’s Verifiable Intent uses SD-JWT delegation chains with eight machine-verifiable constraint types, from amount bounds to merchant allowlists. Vouched’s KYA-OS integrates via REST API alongside existing fraud infrastructure, with an open specification donated to the Decentralized Identity Foundation.
This is not a matter of picking one and moving on. In 2025 alone, eight competing protocols were released – UCP, ACP, AP2, x402, and others. Dual-protocol merchants capture approximately 40% more agentic traffic than single-protocol merchants, which means businesses are effectively forced to build against multiple vendor-specific APIs, MCP servers, or web protocol primitives just to remain visible to agent-driven shoppers.
Integration costs scale aggressively with complexity. A Shopify storefront may see near-zero capital expenditure because the platform handles protocol integration. For merchants using Stripe with custom builds, the range is $5,000 to $50,000. Independent merchants with custom stacks face $50,000 to $200,000. Enterprise-level custom payment service providers are looking at $150,000 to $500,000 or more. The Mastercard and Cloudflare partnership, which lets merchants accept agent transactions without writing custom code through automated trusted-agent filtering, hints at a middleware layer that could eventually lower these costs – but it is a single-vendor solution, not an industry standard.
This fragmentation tax is accumulating against a backdrop of consumer hesitation that should give merchants pause. Checkout.com’s June 2026 survey of 12,000-plus consumers and 400-plus Heads of Payments found 89% of merchants actively preparing for agentic commerce and 42% already testing it. But only 3% of US and UK transactions currently involve AI agents. Product.ai’s April 2026 Trust in AI Commerce Report found that only 14% of US online shoppers trust AI recommendations without verifying them through another source – 86% check before buying. Forty-two percent will not trust AI for purchases over $25. AI tools rank sixth or seventh among information sources, scoring 4.51 out of 10 on trust.
Seventy-two percent of UK and US merchants believe consumers will adopt AI-powered shopping faster than businesses are prepared for. The structural tension is different: merchants are being asked to invest heavily in infrastructure for a transaction layer that remains conditional on consumer trust at a level the numbers do not yet support.
Until the merchant integration burden drops or consumer trust rises, the gap between the two is where the adoption trap lives.