Feedback wanted: building a Shopify-based UCP shopping agent with comparison, trust signals, and redirect checkout

Hi all,

We’ve been building a Shopify-based shopping agent around a UCP-style flow and wanted feedback from others working on similar implementations.

The goal is not only product discovery, but helping customers make better buying decisions across merchants: finding the right product, getting relevant recommendations, comparing similar offers, understanding seller context, and deciding whether to buy now or wait.

So far, our implementation supports:

  • conversational product discovery

  • recommendations based on user intent and follow-up questions

  • comparison across sellers on price, features, and availability

  • product-level trust signals

  • merchant-level trust signals

  • merchant website safety checks

  • price tracking / alert-style follow-up

  • redirecting the user to the merchant storefront for checkout

From the merchant side, the goal is to send better-qualified buying intent into the storefront by reducing comparison friction earlier in the journey.

One design decision we made was to augment the Shopify/UCP-provided data with additional external trust and safety signals. In our implementation, the Shopify/UCP data was useful for discovery and merchant/storefront results, but it did not by itself provide all of the buyer-confidence context we wanted, such as external review/reputation context and basic merchant website safety checks.

Because of that, we added an additional trust/safety layer on top of the core commerce flow.

What we have not completed yet is a fully native in-conversation checkout experience. Our current purchase path still redirects users to the merchant storefront to complete checkout.

Implementation context:

  • JSON-RPC 2.0

  • JWT-based auth

  • ES256 signing support

  • multi-language responses

  • current journey: discovery → recommend → compare → evaluate trust/safety → redirect checkout

A few questions for others building in this area:

  1. Are we heading in the right direction by treating recommendations, product trust, merchant trust, and merchant safety checks as part of the shopping flow, instead of limiting the experience to product retrieval only?

  2. Has anyone here already enabled a more complete checkout experience in a UCP-style implementation beyond storefront redirect? If yes, how are you structuring that flow?

  3. Are others also augmenting Shopify/UCP results with external trust or safety signals, or is there a better pattern for handling merchant confidence and purchase decision support?

  4. How are others handling capability negotiation between discovery only, redirect checkout, and future native checkout support?

  5. If your implementation spans web and messaging surfaces, are you reusing one shared backend/tool layer or handling each surface differently?

Mainly looking for implementation feedback, tradeoffs, and lessons learned from others building similar experiences.

Optional reference implementation for context:
https://ucp.txted.ai

Sharing this only to illustrate the current discovery → recommend → compare → redirect checkout flow.

@sunny_kaiwar
I had a play around in this space before (niche shopping agent) and may pick it up again but busy developing other apps for now - some input for what it’s worth

  1. yeah this was a huge, huge issue for me - we built a great prototype, was getting great results sometimes but other times pure sh1t - a store quality filter was clearly needed (we had a relevance filter forked off something else). getting that data though would be hard. I work with a lot of established clients also and when querying for their products I was finding they were often swamped out by junk. problem with filtering out post query is the limits e.g.
    a. get 10 products
    b. filter out 2-3/10 based on relevance (sometimes none)
    c. do an enrichment pass to add all the stuff Shopify doesn’t give (more images, currency etc)
    c. then do a 3rd party api call to filter out the dropshippers
  2. we didn’t get to that - our model would have been ppc based ultimately and we would want to pass them over to storefront
  3. I would be interested to know on this - Google obviously has the data but doesn’t expose it and matching what Shopify give to real data may be a problem. I think Shopify ultimately should expose some quality signals/filters if they expect people to use this.

4-5 not relevant for us

I will revisit this one soon