Shopify's Big Move
Shopify just announced that millions of merchants can now sell inside AI chats... ChatGPT, Copilot, Gemini, Google AI Mode... all managed from the Shopify Admin through Agentic Storefronts.
And they launched an Agentic Plan that lets brands who aren't even on Shopify upload products to Shopify Catalog and become shoppable across these AI surfaces.
This is a massive move. And I think it validates something that many of us in the agentic commerce space have been saying for a while now... this channel is real, and the infrastructure race is on.
But in many of my current conversations, I keep coming across something that I wanted to dig deeper on... because there's a framing that keeps getting repeated across these announcements that I think deserves more scrutiny.
The Consent Question
The framing is: merchant consent.
UCP, ACP, Shopify's Agentic Storefronts... they all position "merchant consent" and "merchant control" as a core design principle. Merchants opt in. Merchants enroll. Merchants activate their storefronts. Merchants upload their catalogs. Merchants implement endpoints. Merchants approve which agents can transact with them.
And on the surface, that sounds right. Of course merchants should have control over how their products are sold.
But I want to ask a genuine question... is that actually how commerce has ever worked?
How Commerce Has Always Worked
Think about it.
A customer walks into a store and buys something. The merchant doesn't ask who told them about the product. A customer finds a product through Google, clicks through, and checks out. The merchant doesn't gate whether Google is "approved" to send that traffic. Someone reads a blog post, clicks an affiliate link, and purchases. The merchant pays commission but they never pre-approved that specific blog.
The entire open web has operated on a simple principle: if a customer shows up at your checkout with money, and you have the product in stock, the transaction happens. The merchant doesn't gate the demand channel... they gate the transaction itself. Is the payment valid? Is it fraud? Is the item available?
So why is agentic commerce being designed differently?
If an AI agent surfaces a product to a user, and that user wants to buy it, and they have a valid payment method, and the merchant has the item in stock... why does the merchant need to have pre-approved the agent? The order is real. The payment is real. The merchant fulfills it exactly as they would any other web order.
I'm not saying consent doesn't matter at all. There are real cases where merchant control is important... MAP pricing policies, regulated goods, regional restrictions, fulfillment constraints, etc. But those are already enforced at checkout. Every merchant's existing checkout flows already handles these business rules. The transaction itself is the gating mechanism, not enrolling in a protocol...
Who Does Consent Actually Benefit?
So what is the "consent" framing actually doing?
I think it's worth asking who benefits most from this design pattern.
Shopify's Agentic Storefronts require merchants to activate through Shopify Admin. That keeps Shopify in the middle of every transaction. If any agent could buy from any Shopify merchant without going through Agentic Storefronts, Shopify loses its position in that flow. ACP requires merchant enrollment through Stripe. Merchants who don't enroll aren't buyable. That creates a coverage advantage for the platforms that have the most enrolled merchants.
The "consent" framing makes this sound like it's protecting the merchant. And maybe in part it is. But it's also creating a new gatekeeping layer that didn't exist before in commerce, and calling it "merchant empowerment."
"Open" Doesn't Mean "Accessible"
These protocols are open standards. UCP is open source. ACP is Apache 2.0. Any developer can technically build against them. That's great.
But "open" doesn't mean "accessible."
Even with open protocols, the merchant still has to actively implement endpoints. The merchant can reject agents on a per-agent, per-transaction basis. Getting products into ChatGPT via ACP requires merchants to apply and be approved. And if a developer builds a fashion app and wants to sell a product from a DTC brand that hasn't enrolled in any protocol... the protocol is useless.
The Long Tail Gets Left Behind
So who does this leave behind?
The long tail. The millions of merchants on custom platforms, legacy systems, or smaller ecommerce stacks who will never implement UCP or ACP. The thousands of developers building AI-native shopping experiences, gifting platforms, browser extensions, corporate procurement tools, and vertical agents who can't wait for merchant enrollment and don't have the leverage to negotiate protocol implementations with brands.
These developers don't need merchants to adopt a new protocol. They need infrastructure that works with the checkout flows merchants already have.
What We're Building at Rye
That's exactly what we're building at Rye! (absolutely shameless plug)
Our Universal Checkout API lets any developer buy from any merchant using a product URL and a tokenized payment method. No merchant integration. No protocol adoption. No enrollment. No approval. The merchant's existing checkout is the interface, as it always has been. Their business rules, pricing, inventory, shipping, all of it is respected because the order goes through their actual infrastructure, as they have already built.
And if a merchant wants to go further, they can whitelist our RyeBot for better reliability and lower latency. But that's an optimization, not a requirement. The default is zero lift.
Compare that to Shopify's Agentic Plan, which is "free" but still requires brands to sign up, upload product data, manage a Shopify Catalog, install the Knowledge Base app, and optimize structured data for AI readability. That's not zero lift. For a brand on a custom platform or a legacy system, that's a meaningful project.
Free is not frictionless.
Platforms vs. Developers
I think the real distinction is this...
Shopify, UCP, and ACP are building the platform layer of agentic commerce. They're connecting merchants to the top 5-10 AI surfaces... ChatGPT, Copilot, Gemini, Google AI Mode. Those are important surfaces, and this work matters.
But commerce doesn't just happen on five platforms.
The real explosion will be the thousands of apps, agents, and tools that developers are building right now. A fashion app that wants to let users buy any product they discover. A gifting platform that needs checkout across hundreds of brands. A corporate procurement tool automating purchasing. A personal finance app that buys when prices drop. A vertical AI agent for home renovation or fitness or pet care that completes purchases on behalf of users.
None of those developers will have the leverage to get UCP implementations from merchants. None of them will be big enough for Shopify to build a dedicated integration for. And merchants can reject their transactions on a per-agent basis even if the protocol is technically implemented.
These developers need what Rye provides. And every developer building on Rye adds another surface where merchants' products are buyable, which makes the ecosystem more valuable for everyone.
UCP and ACP will serve the top surfaces. Rye serves the next 10,000.
Protocols Aren't Wrong...
I want to be clear... I'm not saying protocols are wrong. I think they represent an important evolution and I'm genuinely excited to see the momentum. We need standards. We need interoperability.
But I also think we need to be honest about what the "consent" model optimizes for, and what it leaves out.
The open web already solved the question of whether a customer can buy from a merchant. The answer has always been yes... if the payment is valid and the product is available, the transaction happens. No intermediary approval needed.
Agentic commerce should work the same way.
If a customer wants to buy, and a merchant wants to sell, the transaction should happen. No gatekeepers. No enrollment. No approval queues. The merchant's existing checkout already enforces their business rules. We just make it accessible to any developer, any agent, any surface.
That's the future we're building toward.
And with our MCP server for universal checkout currently in development, we're making sure that as the protocol ecosystem evolves, Rye is there to fill the gap that protocols will always have... universal coverage for the open web.
Source
*posted originally on LinkedIn - link*
Frequently Asked Questions
How does merchant consent in agentic commerce differ from how commerce traditionally works?
Traditionally, if a customer finds a product through any channel and arrives at checkout with valid payment while the item is in stock, the transaction happens. The merchant gates the transaction itself, not the demand channel. Protocols like UCP and ACP change that by requiring merchants to enroll, approve agents, and implement endpoints before a transaction can occur.
Why do developers building outside the major protocols rely on infrastructure like Rye?
Thousands of developers are building shopping experiences, from fashion apps to gifting platforms to corporate procurement tools, that lack the leverage to negotiate protocol implementations with individual brands. They need infrastructure that works directly with the checkout flows merchants already have, rather than another enrollment requirement. A protocol-only approach leaves those smaller surfaces, and the merchants who have not adopted them, inaccessible.