Shopify MCP Server Raises More Questions Than It Answers
By releasing an MCP server quickly, Shopify seems to be trying to get ahead of providers that might try and create their own Shopify MCP servers and disintermediate Shopify's own control of various agents.
* Storefront allows searches, queries, and checkout against an individual store.
* Global Catalog allows searches across all merchants by trusted agents like Perplexity.
Storefront seems like it is designed for companies like Gorgias and others, etc., to build agents that look and act like Amazon Rufus. Given this is enabled store by store, it seems like the implications are less.
It's the Global Catalog which has wider implications.
In the tame version, this would just obviate the need for any Shopify merchant to use a third-party to provide a data feed and trusted order integration to the LLMs. Which, hey, one less app to buy.
All this is happening so quickly, even Shopify admitted at Editions that they are not sure where this is headed. So, there could be complications.
Having been at ChannelAdvisor for many years in my career, there are complex business rules that often determine what products you want to appear where, how data is displayed, which products to exclude, and how to calculate pricing.
In short, all the functions that channel management software already provide to Amazon today, you might also want for the LLM connection too.
Will this be possible?
Separately, will Shopify allow other MCP servers to automate Shopify also? Will there be a reason to? Or is this a Shopify exclusive function?
Will individual storefronts be able to override the "trusted integration"?
If there is a bug in the trusted integration providing incorrect information, will a merchant be able to disable themselves?
How will trust be incorporated? What will be the process for incorporating trust signals at a storefront, category, or SKU-level? How will that interact with data the LLMs might scrape directly?
How will agent-based price quotes work in scenarios where pricing is not so straightforward?
Will agents take into account promotions and loyalty programs, and how?
What if the Global Catalog data disagrees with other data the LLM scrapes? Who wins?
Many questions, few answers. It's still early but it feels like we are building a messy marketplace here in real-time with few guardrails for merchants.
In reality, a lot of this is on the LLM themselves.
Who will be the first LLM to release a Brand Registry MCP and A2A (naturally) for brands to police how they show up to agents?
Who will be the first LLM to release a data clean-room that brands can access which might contain product/customer data which is being presented to both users and agent interaction? Will "agentic funnel spying" be possible?
Some of these questions could be asked of normal marketplaces, and the line between a scaled global marketplace like Walmart and Amazon and like an OpenAI could blur.
