A sentence on X is rewriting the rules of online commerce: “Instead of navigating HTML built for humans.” With those words, Gil Greenberg, Staff Product Manager for Agentic Commerce at Shopify, described one of the most far-reaching changes in e-commerce in years on September 28 and 29, 2026. Shopify is opening its checkout to AI agents, turning agentic checkout into a feature that no longer reserves the final purchase step for humans alone but also allows automated assistants to complete it, as long as the buyer approves at the end.
While other major platforms are locking AI bots out of their stores, Shopify is moving in the opposite direction. This is not a minor product announcement, but a strategic bet on the future of digital shopping.
Table of Contents
Key Takeaways
- On September 28/29, 2026, Shopify rolled out WebMCP support for the complete checkout flow, including Shop Pay.
- Three new tools, get_checkout, update_checkout and complete_checkout, let agents read, adjust, and complete a checkout.
- Finalizing a purchase still requires explicit authorization from the human buyer; fully autonomous purchases without approval are not part of the design.
- The technology is built on WebMCP and Shopify’s Universal Commerce Protocol (UCP); a hosted MCP server also enables server-to-server communication.
- Amazon and, reportedly, Adidas currently block AI agents from automated checkout, positioning Shopify as a deliberate counterpoint.
What Exactly Did Shopify Announce
On September 28 and 29, 2026, both TechCrunch and Shopify’s official developer documentation confirmed a new capability tier for agentic checkout on Shopify: browser-based AI agents can now not only search products and fill carts, but complete the entire checkout process independently, including payment processing via Shop Pay.
This is a qualitative leap. Previously, agents could already browse storefronts and add items to a cart via WebMCP. The actual payment step remained reserved for humans, because it involves sensitive data such as shipping addresses, payment information, and legally relevant confirmations. Shopify is now closing exactly that gap.
Important for merchants and developers: the feature is rolling out gradually. Not every Shopify store gets the capability immediately — Shopify is enabling it first for eligible merchants before wider availability follows.
The Three New Tools in Detail
At the center of the new capability are three clearly defined tools that agents can call via WebMCP:
| Tool | Function | Typical use |
|---|---|---|
| get_checkout | Reads the current state of the checkout screen | Agent checks prices, shipping costs, available options |
| update_checkout | Changes details such as shipping address or delivery method | Agent adjusts order details at the user’s request |
| complete_checkout | Finalizes the order after authorization | Final purchase confirmation by the buyer |
This structure is deliberately granular. The agent must first understand the current state (get_checkout), can then make adjustments (update_checkout), and is only allowed to actually move money in the final step (complete_checkout). This three-step split creates checkpoints where human oversight can intervene before a transaction becomes final.
For a look at how granular access control for AI agents can be implemented technically, see our article on Docker Cloud Sandboxes for AI Agents.
WebMCP Explained
For developers not yet familiar with the term: WebMCP stands for Web Model Context Protocol. It is a proposed standard that allows AI agents to interact directly with websites inside the user’s browser — not via screenshots, not by scraping HTML, but through structured, machine-readable interfaces.
The difference can be summarized simply:
- Classic approach: An AI agent would have to “see” a webpage like a human, recognize buttons through image analysis, and guess at form fields. That’s error-prone and slow.
- WebMCP approach: The website provides the agent with defined functions, for example “retrieve current shipping address” or “complete order.” The agent calls these functions directly.
Gil Greenberg put it succinctly in his X post: WebMCP tools are “purposely designed, via UCP, to ensure accurate commerce facts, required disclosures, and handoff requirements.” In other words, the interfaces guarantee that the agent receives correct pricing, mandatory disclosures, and the necessary handoff points to the human, rather than having to piece that information together from an interface built for people.
Universal Commerce Protocol as the Foundation
Behind WebMCP sits a larger system at Shopify: the Universal Commerce Protocol, or UCP. UCP is the unified method by which Shopify makes product discovery, cart-building, and checkout accessible to agents, regardless of which AI assistant is operating in the background.
For technically minded readers, UCP is essentially the contract layer that defines how an agent is allowed to “talk” to a Shopify store. In addition, Shopify offers a hosted Model Context Protocol (MCP) server. This is aimed at agents that don’t run inside the user’s browser but operate as a server-to-server connection — for example, when a third-party AI service processes orders on behalf of users in the background.
This split — WebMCP for browser-based interaction, a hosted MCP server for server-to-server scenarios — covers the two most important technical architectures currently being discussed in the agentic commerce space.
Human Control Remains Mandatory
One of the most important clarifications concerns the question of autonomy. Shopify’s agentic checkout is not a system for fully autonomous purchases without human involvement. The complete_checkout tool explicitly requires authorization from the buyer before an order is actually placed.
“Purposely designed, via UCP, to ensure accurate commerce facts, required disclosures, and handoff requirements.” — Gil Greenberg, Staff Product Manager Agentic Commerce, Shopify
That statement is notable in the context of the broader debate around AI safety. Shopify is positioning the technology not as a replacement for human decisions, but as an efficiency tool that shortens the path to a purchase decision — while the final decision still rests with the human.
Existing Partnerships Already in Place
The rollout is not an isolated announcement. Shopify already maintains partnerships in the agentic shopping space:
- Meta’s Muse agent, which enables AI-powered shopping within Meta’s ecosystem.
- Instinct, a partnership announced on the same day as the WebMCP checkout rollout.
These partnerships show that Shopify isn’t just building the technical infrastructure — it is actively pursuing alliances with companies that operate AI agents for end users. For Shopify merchants, that potentially means additional sales channels through which customers might shop in the future, mediated by an AI assistant instead of a traditional web browser visit.
Agentic Commerce as the Bigger Trend
Shopify’s agentic checkout is one building block in a much larger shift the industry calls agentic commerce. The core idea: AI agents are increasingly taking over tasks that humans previously handled manually in the browser, from product research and price comparison to the actual purchase.
Until now, checkout was the last bastion protected from automation. Payment data, legally relevant confirmations, and the emotional moment of the purchase decision were considered too sensitive for bots. With this new feature, Shopify is shifting that boundary significantly.
For e-commerce operators, this means: anyone running a Shopify store could, going forward, receive a growing share of orders not from humans directly, but from their AI assistants. That opens up new revenue potential, since purchase flows could become faster and lower-friction — an agent that already knows all of a user’s preferences doesn’t need to navigate a full product page from scratch.
The Counter-Trend: Amazon and Adidas Lock Agents Out
Not every major platform shares Shopify’s enthusiasm for autonomous buyers. According to consistent reports, Amazon and, apparently, Adidas are currently actively blocking AI agents from automated checkout on their respective platforms.
The reasons are plausible, even if they aren’t spelled out in detail in the available sources:
- Control over the customer relationship: Opening checkout to bots may mean losing direct visibility into purchase decisions and customer data.
- Fraud and abuse risk: Automated purchases could be exploited for price scraping, bot hoarding, or other unwanted patterns.
- Brand control: Platforms with a strong brand experience may worry that agents bypass a carefully designed purchase journey.
Shopify is deliberately choosing the opposite path here. The company is betting that accessibility for agents is a competitive advantage, precisely because other major platforms remain cautious. For Shopify merchants, that could mean in the short term: early adopters benefit from a niche with less competition from other cross-platform offerings.
Opportunities for Shopify Merchants
For Shopify store operators, several concrete advantages emerge once the feature becomes available for their store:
- Lower cart abandonment: When an agent completes checkout on behalf of the user, many classic abandonment causes — cumbersome forms, slow load times — fall away.
- New sales channels: Partnerships such as those with Meta’s Muse agent or Instinct open access to users who shop primarily through AI assistants.
- Early competitive advantage: Since Amazon and Adidas are going the opposite route, Shopify merchants could benefit from users specifically searching for agent-friendly stores.
- More efficient repeat purchases: For recurring orders with known preferences, an agent can significantly speed up the process.
Open Questions and Risks
Despite the clear safeguard of human authorization, open questions remain that the industry will likely debate in the coming months:
- Transparency: Does the buyer always clearly understand which decisions the agent has already made in the background before the final confirmation is requested?
- Payment data security: How is sensitive information such as Shop Pay access protected against misuse by faulty or manipulated agents?
- Liability for errors: Who is responsible if an agent accidentally selects the wrong shipping option or enters an incorrect address, even though the buyer formally approved it?
- Market standard vs. fragmentation: WebMCP is a proposed standard, not a universally binding protocol. Whether this approach becomes an industry-wide norm or competing standards emerge remains an open question.
These points show that while Shopify’s agentic checkout is technically already functional, the accompanying questions around trust and control still need to be worked out in parallel.
Conclusion and Next Steps
Shopify’s agentic checkout marks a turning point for online retail. With WebMCP support for get_checkout, update_checkout, and complete_checkout, Shopify becomes one of the first major e-commerce platforms to open the entire purchase flow to AI agents — safeguarded by mandatory human authorization at the final step. While Amazon and Adidas take the opposite path and actively block agents, Shopify is betting on openness as a strategic advantage, backed by existing partnerships with Meta’s Muse agent and Instinct.
For developers and e-commerce operators, several concrete next steps follow:
- Check whether your own Shopify store is already part of the gradual feature rollout.
- Review the official Shopify Developer documentation on WebMCP and UCP in detail to understand the technical requirements.
- Prepare internal customer service and returns processes for a growing share of orders arriving through agents rather than classic browser sessions.
- Keep watching how WebMCP develops as a proposed standard, since the rules of agentic commerce are likely to keep evolving in the coming months.
Shopify merchants who understand the fundamentals of WebMCP and UCP today are better positioned for a future in which autonomous buyers, accompanied by human approval, become a standard part of online retail.
