On 11 November 2026 Shopify deprecated the Storefront MCP tool that AI agents used to look up a store’s policies and FAQs, and replaced it with Ask (ask_business), a tool that returns a written answer drawn from your knowledge base, policies and product information. Developers have until 15 January 2027 to switch. The same day, the old standalone Knowledge Base app goes away for good. If an AI assistant is asked “can I return this after 30 days?” about your store, this is now the pipe the answer comes through, and your knowledge base is what feeds it.
Having worked in eCommerce technology for over two decades, I believe AI agents will significantly change how customers interact with online stores. As CTO of Binary IC, I see Shopify’s Ask tool as an important step towards making store information accessible through AI-powered conversations. However, the real challenge for merchants is not just adopting AI, but ensuring that product details, return policies, FAQs and other business information are accurate and regularly updated. An AI assistant is only as reliable as the information it receives, and incorrect answers can quickly damage customer trust.
Most merchants I speak to nod along when these acronyms come up and then quietly move on. So here is a glossary instead of a news summary: nine terms, what each one means, and why it matters to the person who owns the store.
The glossary
AI shopping agent
An assistant (ChatGPT, Gemini, Claude, Shop and others) that answers shopping questions or buys on a person’s behalf. Why it matters: the shopper may never see your FAQ page. They see the agent’s summary of it.
MCP (Model Context Protocol)
A common way for AI tools to call outside services, a bit like a standard plug socket. Why it matters: because it is standard, one integration on Shopify’s side reaches many agents at once.
Storefront MCP
Shopify’s earlier MCP endpoint for each store, at /api/mcp. Its search_shop_policies_and_faqs tool returned raw policy text and facts. Why it matters: that tool is now deprecated. It keeps working until 15 January 2027, but all new documentation points to Ask.
UCP (Universal Commerce Protocol)
The open standard Shopify built with Google so agents can discover products, ask questions and check out in a consistent way. Its MCP endpoint lives at /api/ucp/mcp on your store’s domain. Why it matters: this is where Shopify is putting its agent features from now on.
Ask (ask_business)
The new UCP tool. An agent sends one buyer question; Ask returns a plain-text answer built from your knowledge base, policies and product information. It answers questions only. It does not change carts, place orders or start returns, and it handles one question per request, with no multi-turn memory. Why it matters: the shopper now gets a written answer, not your policy page pasted in. If the source is vague, the answer will be vague too.
Agent profile
A small declaration every request must carry, saying which agent is asking and which UCP capabilities it supports (for Ask, dev.ucp.common.ask). Why it matters: mostly a developer detail, but it is how Shopify knows who is asking about your store.
Knowledge base
Your store’s set of facts for AI channels, now inside the admin at Sales channels > Agentic > Knowledge base. Shopify drafts facts from your published storefront pages across ten categories, such as company policies, returns and warranty, payment and fulfilment. Nothing drafted reaches agents until you publish it. Why it matters: this is the closest thing you have to editing what AI says about you.
Additional facts
Facts you write yourself, in three shapes: a paragraph the agent can rephrase, a yes/no answer, or a list of tags. The whole knowledge base is capped at 1,000 facts. Why it matters: anything Shopify cannot draft from your pages, such as your order cancellation rule or COD limits, has to go here or it does not exist for the agent.
Standalone Knowledge Base app
The old separate app, retiring on 15 January 2027. Stores with fewer than 1,000 facts had them moved into the admin and published. Stores with more did not, and must re-add what they want to keep. Why it matters: if you had a large set of facts, check that they actually made the move.
What to do before 15 January 2027
If you run the store
- Open Sales channels > Agentic > Knowledge base and read every drafted fact as if you were a sceptical customer.
- Fix your actual policy pages first. The drafts come from them, so a vague returns page produces a vague fact.
- Add the five questions your support team answers most often as additional facts, in plain words with numbers: days, rupees, pin codes.
- Publish only what is true today. A fact about a sale offer that ended last week will be repeated with confidence.
- Put a monthly reminder in the calendar to review it. Policies change; knowledge bases tend to be written once and forgotten.
If you build agents or apps
- Move calls from
search_shop_policies_and_faqstoask_businesson the UCP endpoint. - Send a
metaobject with yourucp-agent.profile, put the question inask.query, and optional context such as country or language inask.context. - Update your parser: the answer arrives as plain text in
result.structuredContent.answer.plain, not as raw facts. Calltools/listfirst to check the store offers Ask to your profile.
Also in the changelog this fortnight
Two smaller items worth knowing. Customer data erasure requests are now processed 5 days after submission instead of 10 (from 10 November), so apps that sync customer data need to keep up. And multi-factor authentication is becoming mandatory on all Shopify partner accounts, which affects every agency and freelancer who works on your store. The IP allowlist deadline I covered last week is here.
My view
The common advice is “AI will read your website and figure it out”. I disagree. AI will read your website and repeat it, including the parts that were written in a hurry five years ago. This change is good news for merchants who are clear and bad news for merchants who are vague. Writing a knowledge base is copywriting for a new kind of reader, one that never skims and never asks a follow-up. I wrote about where this is heading in Shopify, ChatGPT and the new era of commerce; this is one of the first places it touches day-to-day operations.
How Binary helps
At Binary Web Solutions we work with Shopify Plus brands on exactly this kind of plumbing. We can audit your policy pages and knowledge base against your real support tickets, rewrite them so an agent cannot misread them, and move any custom agent integration to UCP before the deadline. More of our Shopify notes are in the Shopify section.
FAQ
Do I have to do anything if I never used the Knowledge Base app?
Yes, at least look. Shopify drafts facts from your storefront pages for every store, and they reach agents once published. Review them under Sales channels > Agentic.
Can Ask process a return or cancel an order?
No. Shopify’s documentation is clear that Ask only answers questions. It does not change carts, place orders or start returns.
What happens on 15 January 2027?
Two things: the deprecated Storefront MCP policies tool stops being maintained, and the standalone Knowledge Base app is no longer available. Facts must live in the admin’s knowledge base by then.
Useful for this change
- QwikAns AI by DiGiMAA: the same idea on your own product pages, so shoppers who are on your site get answers there. Free on the Shopify App Store.
- SupportHub Helpdesk: your complaint log is the best source for the facts your knowledge base is missing.
- Need it built for you? Binary does this for Shopify Plus brands: policy rewrites, knowledge base audits and UCP integrations.
- Ommniverse, my book on online and offline customer experience: why a clear answer is part of the product.
Sources
- Shopify changelog: Storefront MCP policies and FAQs tool deprecated in favour of UCP MCP Ask (11 Nov 2026)
- Shopify docs: Ask
- Shopify changelog: Knowledge base is now available in your admin
- Shopify changelog: customer data erasure requests updated from 10 days to 5
Tejash Shah is Co-Founder and CTO of Binary Web Solutions, a Shopify Plus agency in Mumbai and Ahmedabad, and founder of DiGiMAA.











