Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open protocol developed by Anthropic that allows AI models to securely connect with external data sources and tools, enabling AI agents to query product catalogues, check pricing, and interact with business systems in real time. Think of it as a universal adapter between AI models and the outside world.
What is the Model Context Protocol?
MCP solves a fundamental limitation of large language models: on their own, they can only work with the data they were trained on. They cannot check your current inventory levels, query your product database, or look up today's pricing. MCP changes this by providing a standardised way for AI models to connect to external data sources and tools: securely, in real time, and without requiring custom integrations for each AI system.
The protocol works through a client-server architecture. A business publishes an MCP server that exposes specific capabilities (product catalogue queries, pricing lookups, inventory checks, order status retrieval) through defined tools and resources. Any MCP-compatible AI agent (the client) can then connect to that server and use those capabilities. This standardisation means a manufacturer only needs to build one MCP server, and it works with every MCP-compatible AI agent, regardless of which company built the agent or which language model powers it.
While MCP was developed by Anthropic for its Claude ecosystem, it has been released as an open protocol and is gaining adoption across the broader AI industry. This positions it as a general-purpose connectivity layer for AI, not limited to commerce, but with significant implications for how AI procurement agents access supplier data and how businesses participate in agentic commerce.
Why MCP Matters for B2B Businesses
For manufacturers and B2B suppliers, MCP represents the second-priority protocol after UCP for agentic commerce readiness. The Claude ecosystem is growing rapidly, with enterprises increasingly deploying Claude-based agents for procurement, research, and operational tasks. By publishing an MCP server, a manufacturer makes its product catalogue, pricing, and availability data directly accessible to these agents, without requiring any intermediary or marketplace.
MCP's general-purpose nature is both its strength and its distinction from commerce-specific protocols. Where UCP handles the full purchasing transaction and ACP manages payment processing, MCP excels at data access and tool interaction. An AI procurement agent might use MCP to query a supplier's product catalogue, check real-time inventory, download technical specifications, and verify certifications, then hand off to UCP or ACP for the actual purchase. This makes MCP complementary to the commerce protocols, not a replacement for them.
The practical advantage for manufacturers is flexibility. An MCP server can expose whatever data and capabilities the business chooses, from simple product lookups to complex configuration tools for custom-manufactured items. A manufacturer of industrial valves, for example, could publish an MCP server that lets AI agents specify flow rates, pressure ratings, and material requirements, then returns matching products with accurate pricing. This level of interactive product discovery goes beyond what static data feeds can offer and creates a significant advantage in agent-driven procurement workflows.
How MCP Works
MCP uses a client-server architecture with three core primitives: tools, resources, and prompts. Tools are functions that the AI agent can call, for example, "search_products", "get_pricing", or "check_inventory". Resources are data that the server exposes for the agent to read, such as product catalogues, technical documentation, and compliance certificates. Prompts are predefined interaction templates that guide the agent in using the server's capabilities effectively.
A business implements an MCP server by defining which tools and resources to expose and connecting them to its existing systems, such as ERP, PIM (product information management), inventory management, or pricing engines. The server handles authentication, rate limiting, and data formatting, presenting a clean interface that any MCP client can consume. Importantly, the business retains full control over what data is exposed and what actions agents can take.
When an AI agent connects to an MCP server, it first discovers the available tools and resources through a capability negotiation step. The agent then uses these capabilities as needed within its workflow. For a procurement scenario, this might involve searching for products matching specific criteria, retrieving detailed specifications for shortlisted items, checking availability and lead times, and requesting a quote, all through standardised MCP calls that work identically regardless of which AI system is making the request.
Key Takeaways
- MCP is Anthropic's open protocol for connecting AI models to external data sources and tools; a universal adapter for AI-to-business-system communication.
- Second priority after UCP for manufacturers preparing for agentic commerce, driven by rapid growth of the Claude enterprise ecosystem.
- Complementary to commerce protocols (UCP, ACP): MCP handles data access and tool interaction, while commerce protocols handle transactions.
- MCP servers give manufacturers full control over what data and capabilities are exposed to AI agents.
- Enables interactive product discovery that goes beyond static data feeds, particularly valuable for configurable or custom-manufactured products.