I found this topic that was pointed here: Feature Request: Improve Shopify MCP Agent to Use Metafields for Structured Product Data
Cant seem to find a post here about it. The OP is correct. If you have metafield it is not searched. This can be demonstrated with the demo snowboard store. If you query for a length, color, etc…It returns 0 results.
Hey @Mike_G , right now, in general for our MCP tools, metafields are currently unsupported, but I can definitely look into confirming this specifically when it comes to product searches. Out of curiosity, is this popping up when you the AI agent runs the search_shop_catalog tool (more info on this here)?
Hope to hear from you soon.
Hi @Alan_G ! Yes, that is the process I am using and ‘search_shop_catalog’ tool being used. Im sure its the same deal, but i have noticed the same thing from the MCP Shopify connector used in OpenAI’s Agent Builder as well.
I hope this gets updated. Right now we are using GraphQL to query all the product info, stuffing it in a vector store, and implementing our own “search_store” tool which is not optimal, but works.
Thanks @Mike_G - appreciate you clarifying. Yeah, right now this is a limitation, but I’ll look into things on my end here for you and loop back if I’m able to share any information (or confirm that I’ve set up a feature request for you at the very least). Speak with you soon!
Hey @Mike_G - circling back here. At the moment, I don’t have more info to share on metafield integration on my end (in terms of a timeline, etc), but I’m going to set up a feature request for you. If you’re open to sharing a few more details on your use case that would be super appreciated so I can advocate a bit better on my end, hope to hear from you soon!
Hi @Alan_G, no problem. We first notices this in testing with the demo snowboard store. Our goal was to query the MCP and return a structured response of products that fit. This works for basic product questions such as: is ‘X’ snowboard in stock? What is the most expensive snowboard you sell?
But the demo store also has metafields to describe the features of the snowboard such as length, color, binding type, and generally anything specific to a type/brand/model of snowboard. So for instance the queries “Do you have any snowboards over 160cm?”, “Do you have any blue snowboards?”, or “Do you have any snowboards with optimistic bindings?” will all fail even though there are snowboards available with those properties.
This is to help our (chatbeacon.io) customers that use Shopify for their site, provide a product search/add-to-cart functionality. Attached is a screenshot of our integration.
Hey @Mike_G - thanks for sharing the use case. The snowboard store example with specific metafield queries (length, color, etc) is super helpful, definitely get where you’re coming from on this (would be much cleaner to have this data searchable directly through the MCP tools). I’ve logged a feature request with your use case details. I appreciate you sharing the workaround you’re using as well (GraphQL + vector store - guessing this is a separate vector DB you’re using?). I’ve set up that feature request on my end, but let me know if I can help out with anything else.
Thanks again for taking the time to share your use case, not ideal that this isn’t available yet, but I do appreciate you taking the time to share so we can hopefully improve things in the future (especially making our agentic frameworks a lot more robust).
Just going to close this thread out for now, let me know if I can help out further @Mike_G
