Model Context Protocol (MCP)

Definition

The Model Context Protocol (MCP) is an open standard created by Anthropic that defines how AI models connect to external tools, data sources, and services. Think of it as a universal adapter - instead of building custom integrations for every AI model and every service, MCP provides a single protocol that any AI agent can use to interact with any compatible system.

MCP follows a client-server architecture. The AI model acts as the client, sending structured requests. The server exposes capabilities - tools to call, data to read, prompts to use. The protocol handles the handshake, capability discovery, and message exchange between both sides.

For ecommerce, MCP is the foundational layer that makes agentic commerce possible. When Shopify launched its product feed through MCP, it gave AI agents like Claude a structured way to browse, search, and retrieve product data from Shopify stores. Without MCP, each AI provider would need to build bespoke integrations with each ecommerce platform.

Why It Matters

MCP is shifting how products get discovered online. In the traditional web, search engines crawl HTML pages. In the AI-native web, agents query structured data through protocols like MCP. If your store exposes product data through MCP, AI agents can find your products. If it does not, those agents are blind to your catalog.

For merchants, this has practical implications. Shopify stores with MCP-compatible feeds are already surfacing in AI conversations on Claude and other LLM-based assistants. The protocol handles not just product search but also filtering by price, availability, category, and other attributes - giving AI agents the same capabilities a human shopper would have on your website.

The broader significance is interoperability. MCP is model-agnostic. A server built for Claude works with any MCP-compatible client. This means merchants who invest in MCP compatibility are not locked into a single AI provider. They are building for an ecosystem.

How It Works

MCP operates through three core primitives:

Tools are functions the AI model can call. In ecommerce, a tool might be “search_products” or “get_product_details.” The server defines what tools are available, what parameters they accept, and what they return.

Resources are data the server exposes for the AI to read. Product catalogs, inventory levels, store policies - anything the AI might need as context to help a customer.

Prompts are templates the server provides to guide how the AI should use the available tools and resources for specific tasks.

A typical flow works like this: an AI agent discovers an MCP server, queries its available capabilities, then calls tools and reads resources as needed to fulfill a user request. If someone asks “find me a blue wool sweater under $100,” the agent uses MCP tools to search the store’s catalog, filter results, and return relevant products with accurate pricing and availability.

The protocol uses JSON-RPC 2.0 for message formatting and supports multiple transport layers including HTTP with Server-Sent Events. This makes it lightweight enough for real-time interactions while remaining robust enough for complex multi-step workflows.

For merchants on platforms that support MCP (Shopify being the first major one), the setup is largely handled by the platform. The merchant’s job is ensuring their product data - titles, descriptions, images, attributes - is rich and accurate enough to perform well when an AI agent queries it.

Related Terms

• Agentic Commerce Protocol (ACP) - Shopify’s protocol built on MCP specifically for commerce transactions
• Agent-to-Agent Protocol (A2A) - Google’s protocol for AI agents to communicate with each other
• Tool Use - The AI capability that MCP enables and standardizes
• Agentic Commerce - The broader shift that MCP infrastructure supports