Definition: what is MCP?
The Model Context Protocol was published by Anthropic in November 2024 as an open standard and defines a client-server architecture based on JSON-RPC: an MCP server exposes resources (data), tools (actions) and prompts; an MCP client — the AI system — discovers and uses them through the standardised protocol. Instead of building separate integrations per model, a service implements MCP once and becomes connectable for all compatible systems.
The standard spread quickly beyond the Anthropic context; other major AI providers have built support as well. MCP has thus become the leading candidate for the role HTTP played for the web: the connection protocol between AI agents and the world of services.
How MCP works
- Resources: structured data the server exposes — documents, database extracts, product catalogues.
- Tools: executable actions with defined parameters — search, booking, calculation, status queries.
- Prompts: predefined task templates the server offers the client.
The client discovers the server's abilities at runtime (capability discovery) — an agent does not need to know in advance what a service can do; it asks. Exactly this discoverability makes the protocol scalable for the agentic web.
Why MCP matters for visibility strategy
Search surfaces are shifting from “reading answers” to “completing tasks”: AI agents increasingly research, compare and act autonomously. In that world a brand has three presence layers: as content (citable sources, llms.txt), as an entity (knowledge graphs, schema) — and as a service agents can use directly. MCP is the standard of the third layer. In commerce contexts it is complemented by payment and transaction protocols such as x402 and the Agentic Commerce Protocol (ACP).
Practically: exposing product data, availability or booking functions as an MCP server makes them directly usable for agents — instead of hoping an agent scrapes and operates your website correctly.
Related terms
MCP is a core signal of agent readiness — alongside crawler accessibility, structured data and checkout capability. The commerce side of the agentic web is covered in Agentic commerce & SEO.
MCP is connectability to the agentic web
AI agents complete tasks, not just answers. MCP defines how they connect services and data — brands that are not connectable there are just scrapeable text to agents.
FAQ on Model Context Protocol
What is the Model Context Protocol? ▾
An open standard (published by Anthropic in November 2024) defining how AI systems connect external tools, data sources and services — through a client-server architecture based on JSON-RPC with resources, tools and prompts. It has spread beyond the Anthropic context into an industry standard.
Why would a website need MCP? ▾
For the service layer of the agentic web: AI agents can use MCP servers directly — query product data, check availability, execute actions — instead of scraping the website. For transactional business models this becomes an infrastructure question.
How does MCP differ from a classic API? ▾
MCP standardises what is bespoke in classic APIs: discovery, description and call conventions. An agent can use any MCP server without prior knowledge because capabilities are queried at runtime — classic APIs require custom integration each time.
How do MCP, x402 and ACP relate? ▾
MCP connects agents with tools and data; x402 and the Agentic Commerce Protocol address payment and transaction handling of agentic purchases. Together they form the protocol stack through which agents not only read but transact.