case studyShopifyAI chatbotour producte-commerce
🛍️

MCPify:
AI Assistant for Shopify Stores — Our Case Study

· 8 min read · Alexey Mikhailov

Bottom line: Shopify stores lose sales due to slow support — 70% of buyers leave if they do not get an answer within 5 minutes. We built MCPify: an AI chatbot that connects to any store in 5 minutes via a script tag, knows the full catalog and order history, and replies in 1 second 24/7. On our test store it resolved 78% of questions without human involvement.

< 1 sec
average response time
vs 4–8 hours for a human
78%
questions resolved without human
test store, 30 days
5 min
setup time
script tag → works immediately
24/7
availability
including weekends and holidays

What problem were we solving?

A Shopify store owner receives dozens of identical questions every day: "Where is my order?", "What size should I get?", "How do I return an item?", "Is the blue color in stock?". Each question requires a manual reply — or hiring a support agent. For small and mid-sized e-commerce that is either hours of the owner's personal time or $800–2,000/month in payroll.

Existing solutions — Tidio, Gorgias, Zendesk — are either too expensive ($300–1,500/month), or require manual knowledge base setup that goes stale with every catalog update. Real-time integration with actual store data (orders, inventory, SKUs) is a paid add-on with most of them — and still technically complex.

We wanted to do it differently: one script tag → AI immediately knows the full store via Shopify API → it works. No manual knowledge base filling, no technical knowledge required from the store owner, no month-long onboarding.

What can MCPify do?

📦
Order status
Buyer enters order number → AI replies in real time via Shopify API
🔍
Product questions
Ingredients, sizes, availability, comparisons — AI knows the full catalog
📋
Store policy
Returns, shipping, warranty — from the store knowledge base
🛒
Purchase assistance
AI asks clarifying questions and recommends the right product
🌍
Multilingual
Replies in the language of the question: EN, RU, ES, DE, FR
5-minute setup
One script tag in the Shopify theme — nothing else required

How does the architecture work?

MCPify runs on Cloudflare Workers — edge infrastructure with zero cold start and global distribution. Each store gets an isolated context in Cloudflare KV: product data, orders, and store policy synced through the Shopify Admin API. The AI model — Llama 3.3 70B via Groq — processes the customer question, retrieves relevant data from KV, and generates a response in 400–800 ms.

Tech stack
Runtime
Cloudflare Workers
Edge deployment, 0 ms cold start, globally distributed
AI model
Groq + Llama 3.3 70B
Response in 400–800 ms, 14,400 req/day free tier
Storage
Cloudflare KV
Store data: products, orders, FAQ per isolated store
Integration
Shopify Admin API
Live order status, inventory, product data
Widget
Vanilla JS + CSS
Script-tag injection, zero dependencies
Auth
HMAC + store token
Security: each store runs in fully isolated context

The widget is injected into the Shopify theme via a standard script tag — no App Bridge, no OAuth flow for the buyer, no changes to the theme code. The store owner adds one line in the theme settings and the widget appears on all store pages. Store authentication uses an HMAC-signed token; each store runs in a completely isolated context.

Results on the test store

Live Shopify store, 30 days of MCPify operation
Questions received
340 over 30 days
Resolved without human
265 (78%) order status, availability, policy
Routed to human
75 (22%) non-standard situations, complaints
Average AI response time
0.7 sec vs 3–6 hours manually
Setup time
4 minutes from script tag to first reply

78% of questions resolved without a human — meaning the store owner saved ~265 manual replies in one month. At 5 minutes per reply, that is 22 hours of working time saved per month.

MCPify Pricing

Starter
$99/mo
500 replies/mo
Order status
FAQ bot
English support
Most popular
Growth
$249/mo
2,000 replies/mo
Full catalog
Multilingual
Analytics
Scale
$499/mo
Unlimited replies
Custom prompt
Priority support
White-label

Try MCPify on your Shopify store

Setup takes 5 minutes. First customer reply in 10 seconds. 14-day free trial.

Open MCPify →

Want a similar solution for your business?

MCPify is one example of how we build AI systems for specific business problems. If you run an e-commerce store, online school, agency, or any business with a high volume of repetitive requests — we will do a free audit and show how AI changes the economics of your support operation.

Get a free audit →

Frequently Asked Questions

What is MCPify?

MCPify is a plug-and-play AI chat assistant for Shopify stores. It connects to your store in 5 minutes via a single script tag, reads your full product catalog and order history through the Shopify Admin API, and answers customer questions instantly — 24/7. No manual knowledge base setup, no developer required.

How is MCPify different from Tidio or Gorgias?

Tidio and Gorgias require manual knowledge base setup that goes stale every time your catalog updates. They also cost $300–1,500/month and do not connect to real-time order data out of the box. MCPify reads your live Shopify data automatically — products, inventory, order status — and updates in real time. Pricing starts at $99/month.

What percentage of questions does MCPify resolve without a human?

In our 30-day test on a live Shopify store, MCPify resolved 78% of 340 incoming questions without human involvement. These included order status checks, product availability, size questions, and return policy queries. The remaining 22% — non-standard situations and complaints — were routed to the store owner with full conversation context.

How does MCPify handle security?

Each store operates in a fully isolated context in Cloudflare KV. Authentication uses HMAC-signed tokens — no store can access another store's data. The widget is injected via a script tag and communicates only with the store's dedicated Cloudflare Worker instance.

What languages does MCPify support?

MCPify detects the language of the incoming message and responds in the same language. Currently supported: English, Russian, Spanish, German, and French. Multi-language support is included in all plans at no extra charge.