Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open protocol by Anthropic that standardizes communication between AI models and external tools. MCP is not a commerce protocol — but it is the infrastructure layer on which Agentic Commerce is built.

What Is MCP?

MCP solves a fundamental problem in AI development: without a standard, every combination of AI model and external tool would need to be individually integrated. With hundreds of LLMs and thousands of tools, that would not scale. MCP creates a common language.

The USB Metaphor

The most widely used explanation for MCP is the USB analogy:

Before USB, every device had its own connector — printers, mice, keyboards, scanners, each with a different plug. USB standardized the connection: one port for everything. Every USB device works with every USB-capable computer.

MCP does the same for AI: one protocol connects any AI model with any external tool. A store provides an MCP server — and any MCP-compatible agent can use it, whether Claude, ChatGPT, Gemini, or a specialized commerce agent.

MCP in Agentic Commerce

MCP is relevant to Agentic Commerce on two levels:

1. Stores as MCP Servers

Online stores can provide MCP servers that give AI agents access to product data, search functions, and checkout. The agent calls the MCP server and receives structured responses — which products are available, what they cost, how checkout works.

2. Tool Discovery

A particular advantage of MCP over pure REST APIs: Tool Discovery. An MCP server describes its capabilities in a machine-readable format — the agent automatically recognizes which actions are possible. "This store offers product search, cart, and checkout" — without the agent developer needing to read the API documentation.

For merchants, this means: an MCP integration makes your store accessible not just to a specific agent, but to the entire MCP ecosystem.

Shopify MCP Server

The most prominent example of MCP in commerce is the Shopify MCP Server. It gives AI agents access to:

• Product catalog: Search, filtering, product details, variants, prices
• Cart: Create, modify, add/remove products
• Checkout: Complete order and pay
• Order management: Order status, tracking information

For the more than 2 million Shopify merchants, this means: their products are automatically reachable by MCP-compatible agents — without custom development work. This is a significant advantage over stores that must build their own MCP integration.

How an MCP Server Works

An MCP server has three core components:

• Tools: Actions the agent can perform (e.g., "search product," "create cart," "complete checkout"). Each tool has a name, a description, and defined parameters.
• Resources: Data the agent can read (e.g., product catalog, category list, shipping options). Static or dynamic information.
• Prompts: Predefined instructions that help the agent use the tools correctly (e.g., "For clothing, always ask about size").

Communication runs over JSON-RPC. The agent sends a request ("Call tool 'product_search' with parameters {query: 'running shoes', max_price: 120}"), the MCP server processes the request and returns a structured result.

MCP + ACP, MCP + UCP

MCP is complementary to the commerce protocols:

• MCP + ACP: A merchant can expose their ACP Checkout API as an MCP server. The agent discovers the checkout capability automatically via Tool Discovery and uses the ACP endpoints through MCP. This is the approach Shopify pursues.
• MCP + UCP: UCP is explicitly MCP-compatible. Stores can register UCP-compliant endpoints as MCP tools. Google recommends this approach for merchants seeking maximum agent compatibility.

The rule of thumb: ACP and UCP define what an agent can do (product search, checkout, payment). MCP defines how it does it (communication protocol, Tool Discovery, data format).

For merchants, this means: invest in structured data and API endpoints. Whether those are then delivered via ACP, UCP, MCP, or a combination is an implementation detail — data quality is the decisive factor.