Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard for connecting AI applications to external tools and data sources, so an agent can call a function or read a record instead of guessing from training memory. Anthropic introduced it in November 2024 and donated it in December 2025 to the Agentic AI Foundation, a directed fund under the Linux Foundation, which moved it out of single-vendor governance. It is the plumbing layer beneath commerce protocols, not a commerce protocol itself.
What it standardizes
Before MCP, every combination of AI application and data source needed its own integration. MCP defines one wire protocol between a client (the AI application) and a server (whatever exposes the data or capability), which is why its documentation describes it as "a USB-C port for AI applications."
A server exposes three kinds of thing: tools an agent can invoke, resources it can read, and prompts it can use. For a catalog, a search_products tool and a get_product resource are the obvious shapes.
Support is broad rather than vendor-specific. Claude, ChatGPT, VS Code, Cursor and many others act as MCP clients, which is the practical argument for building one server rather than several bespoke integrations.
Governance, and why it matters commercially
Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation, in December 2025, co-founded with Block and OpenAI.
That matters when you are choosing what to integrate. Neutral foundation governance is a reasonable predictor that a protocol will outlive one company's product strategy, which is not something you can say about every standard in this space right now. It is the strongest governance position of the three protocols distributors are currently asked to evaluate — ACP remains OpenAI and Stripe maintained and still labelled beta, and UCP is Google-led.
A catalog over MCP
The appeal for a distributor is real but narrower than the enthusiasm suggests. An MCP server over your catalog lets an agent query live data — current stock at a branch, contract pricing for an authenticated account, the current revision of a spec sheet — rather than working from whatever it scraped last month.
The integration is the easy part. A minimal server exposing product lookup and search is days of work for a competent team. What it exposes is the hard part, and it is the same hard part as everywhere else: if the record lacks the pressure rating, the agent's query returns nothing useful and the protocol has not helped.
There is also an authorization question that commerce teams underestimate. B2B pricing is contract-specific, availability is branch-specific, and some catalog content is customer-specific. An MCP server has to carry identity through, or it will either leak pricing or return a generic answer that is wrong for the buyer asking.
MCP versus the commerce protocols
Three layers, routinely conflated:
| Layer | Scope |
|---|---|
| MCP | How an agent reaches tools and data of any kind |
| ACP | Checkout between a buyer's agent and a merchant |
| UCP | Catalog lookup, cart, identity, checkout, and orders as a merchant surface |
MCP is general-purpose infrastructure that happens to be useful for commerce. UCP explicitly supports MCP alongside REST and JSON-RPC. None of the three improves a thin product record; they determine who can read it and how.
Frequently asked questions
What is the Model Context Protocol?
An open standard for connecting AI applications to external tools and data sources. A server exposes tools, resources and prompts over a defined protocol, and any compatible AI client can use them, which replaces one-off integrations between each model and each data source.
Who governs MCP?
Anthropic introduced it in November 2024 and donated it in December 2025 to the Agentic AI Foundation, a directed fund under the Linux Foundation co-founded with Block and OpenAI. It is now under neutral governance rather than a single vendor's roadmap.
Is MCP a commerce protocol?
No. It is general-purpose infrastructure for connecting agents to tools and data. Commerce-specific protocols sit above it: the Agentic Commerce Protocol covers checkout, and Google's Universal Commerce Protocol covers the wider merchant surface and supports MCP as one of its transports.
Should a distributor build an MCP server for its catalog?
It is worth doing once the underlying record is worth querying, and it solves the freshness problem for stock and pricing that scraped pages cannot. Plan for identity pass-through from the start, since B2B pricing and branch availability are account-specific and a generic answer is often a wrong one.