Model Context Protocol (MCP)

The Model Context Protocol is an open standard originally created by Anthropic that provides a universal way for AI agents to connect to external data sources, tools, and business systems in real time.

Think of MCP as a "USB-C port" for AI — a single, standardized connection that lets any compatible AI agent plug into any compatible data source. Before MCP, every AI integration required custom API development. A retailer connecting their product catalog to ChatGPT needed one integration, connecting to Claude needed another, and each new AI platform meant more engineering. MCP replaces these one-off connections with a universal standard.

Anthropic launched MCP in November 2024. OpenAI adopted it in March 2025. Google DeepMind followed in April 2025. By December 2025, MCP had reached 97 million monthly SDK downloads, over 10,000 active servers, and 300+ client applications (Anthropic, Dec 2025). In December 2025, Anthropic donated MCP to the Linux Foundation's Agentic AI Foundation (AAIF), making it a vendor-neutral, community-governed standard alongside contributions from AWS, Google, Microsoft, and OpenAI. For what the Linux Foundation donation changes for retailers, see Paz.ai's analysis of the MCP donation and Adobe's move to make MCP the default agent protocol for commerce.

MCP is the data connectivity layer that makes agentic commerce possible. While ACP handles checkout and UCP covers the commerce journey, MCP is what lets AI agents actually access your product data, inventory levels, pricing, and order information in real time.

Without MCP (or equivalent connectivity), AI agents are limited to whatever product information was available when the AI model was last trained — which could be months or years out of date. MCP provides live, real-time access to merchant systems.

The scale of adoption is significant. With 97 million monthly SDK downloads and support from every major AI platform (OpenAI, Google, Anthropic, Microsoft), MCP is the most widely adopted protocol in the agentic AI ecosystem. Google's UCP was explicitly designed to be compatible with MCP, using it as one of three integration pathways alongside REST APIs and A2A (Google Developers, Jan 2026).

For commerce specifically, MCP enables AI agents to perform actions that go far beyond product recommendations: checking real-time inventory, applying promotional pricing, processing returns, managing loyalty points, and accessing customer order history (with appropriate permissions). Shopify has implemented MCP servers that let AI assistants manage store operations, product listings, and customer interactions directly (Shopify, 2025).

MCP uses a client-server architecture where AI agents are clients and business systems are servers:

MCP Servers expose specific capabilities — product catalog access, inventory queries, order management, etc. A retailer might run an MCP server that provides real-time access to their product database, inventory management system, and order processing system.

MCP Clients are the AI agents (ChatGPT, Claude, Gemini, etc.) that connect to these servers to access data and perform actions. When a consumer asks ChatGPT, "Is this jacket available in size M?", the AI agent queries the retailer's MCP server for real-time inventory data rather than guessing based on training data.

The protocol handles authentication through OAuth 2.1 (added in the June 2025 specification update), supports both local and remote connections, and includes standardized error handling and capability negotiation.

In practice, a commerce MCP implementation might expose these capabilities:

• Product search: Query the catalog by attributes, category, price range
• Inventory check: Real-time stock availability by size, color, location
• Pricing: Current prices including active promotions and loyalty discounts
• Cart operations: Add/remove items, calculate totals, apply coupons
• Order status: Track shipments, check delivery estimates

MCP is foundational — it sits beneath both ACP and UCP as the data access layer. A retailer can use MCP to expose their catalog, ACP to enable ChatGPT checkout, and UCP to enable Google AI Mode. They are complementary, not competing.

• MCP is the universal connectivity standard for AI agents — the "USB-C" of AI integration
• Adopted by every major AI platform: OpenAI, Google, Anthropic, Microsoft
• 97 million monthly SDK downloads and 10,000+ servers as of December 2025
• Now governed by the Linux Foundation (vendor-neutral, community-driven)
• Essential for agentic commerce — lets AI agents access real-time product data, inventory, and pricing
• UCP and ACP both use MCP as a connectivity option