Skip to content

How to Integrate a Sales Assistant with Shopify MCP

Overview

IMPORTANT: THIS IS A BETA VERSION

It is under active development and may contain unstable or incomplete features. Use it at your own risk. The module behavior may change without notice. Do not use it in production environments.

This guide explains how to configure an intelligent Sales Assistant in Etendo Copilot that connects to a Shopify Model Context Protocol (MCP) server to deliver a guided, conversational buying experience.

The assistant can:

  • Search the product catalog in real time
  • Retrieve detailed product information
  • Manage the shopping cart
  • Access policies and FAQs
  • Provide contextual, consultative recommendations

Prerequisites

  • Etendo Copilot installed and configured
  • At least one working OpenAI / compatible LLM provider configured (or an internal model via Ollama)
  • A functioning Shopify MCP server endpoint exposing tools (HTTP or SSE transport)
  • Network access from the Copilot backend to the Shopify MCP endpoint

Supported Shopify MCP Tools (Examples)

Typical Shopify MCP server implementations expose tools similar to:

Purpose Example Tool Name
Catalog search search_shop_catalog
Product details get_product_details
Cart state get_cart
Update cart update_cart
Policies & FAQs search_shop_policies_and_faqs

Tool names may vary depending on the MCP server implementation. Confirm by inspecting the server's capabilities response.

Step 1. Configure the Shopify MCP Server

Application > Service > Copilot > MCP Servers Configuration

  1. Open the MCP Servers Configuration window (System Administrator role).
  2. Create a new record and paste the JSON configuration.
  3. Save. The system validates and normalizes the structure.

JSON Configuration Examples

Minimal HTTP-based configuration:

{
  "mcp": {
    "servers": {
      "shopify": {
        "url": "https://your-shop.example/api/mcp",
        "transport": "streamable_http"
      }
    }
  }
}

Application > Service > Copilot > Agent

  1. Open (or create) an Agent.
  2. Go to the MCP tab.
  3. Add a new record selecting the Shopify MCP configuration.
  4. Save.

Step 3. Author a Specialized Prompt

Provide clear behavioral instructions so the assistant acts as a consultative digital salesperson.

Suggested base prompt:

You are a professional e-commerce sales assistant for a Shopify-powered store integrated into Etendo.
Objectives:

- Understand the user's intent (need, preference, constraints: price, category, use case).
- Use catalog search BEFORE answering if more context is required.
- When recommending products: list 3–5 options (if available) with: Name | Key benefit | Differentiator.
- When the user asks for differences: respond with a concise comparison table (textual) and a short recommendation.
- Always confirm before adding or updating cart items.
- Use cart tools to reflect real state before confirming.
- If policy / FAQ info is requested, call the policies/faq tool.
- Ask one clarifying question if intent is ambiguous.
- Avoid hallucinating unavailable products.
Tools to consider (call only when needed): search_shop_catalog, get_product_details, get_cart, update_cart, search_shop_policies_and_faqs.
Return final answers in clear, concise business English.

Prompt Refinement

Iterate based on transcripts. Add guardrails for pricing, availability disclaimers, or upsell logic.

Step 4. Test the Assistant

  1. Start a conversation with the configured agent.
  2. Try discovery queries first (e.g. "I need a lightweight running shoe").
  3. Ask for comparisons (e.g. "Compare the first two you suggested").
  4. Add to cart (e.g. "Add the second shoe in size 42"). The agent should call update_cart after confirmation.
  5. Ask for policies (e.g. "What is the return policy?").

Example Conversation

User: "What product do you recommend for home coffee brewing?"

Agent: (Calls search_shop_catalog)

"Here are three options: ... Would you like a comparison or more details on one of them?"

User: "Compare the first two."

Agent: (Calls get_product_details twice)

"Comparison: 1) Model A – Compact | Fast heat | Best for beginners. 2) Model B – Larger capacity | Dual temperature control | Better for experimentation. Recommendation: If you value simplicity choose Model A; for control choose Model B."

User: "Add Model B to my cart."

Agent: "Confirm: add Model B (default configuration) to cart?"

User: "Yes."

Agent: (Calls update_cart, then get_cart)

"Added. Your cart now contains: ..."

This work is licensed under CC BY-SA 2.5 ES by Futit Services S.L..