What is an MCP? Model Context Protocol explained

Learn how the Model Context Protocol (MCP) lets AI and LLM clients connect to external tools, resources, and prompts through a standard server interface.

What is an MCP

The Model Context Protocol (MCP) is a standard interface that allows AI or LLM clients to connect to external tools and data sources through a separate server process. It defines a common way for a client (or host) to discover capabilities, call tools, and access resources without hard‑coding against any specific backend. MCP focuses on the protocol for describing and invoking capabilities so different servers and clients can interoperate cleanly.

• MCP server: a process that exposes tools, resources, and prompts over the protocol
• Client/host: an AI/LLM application or environment that connects to one or more MCP servers
• Tools: callable operations (with typed inputs/outputs) that the client can invoke
• Resources: read‑only or structured data surfaces (files, APIs, datasets) exposed via the server
• Prompts: reusable, parameterizable prompt templates that the client can list and use
• Transports: channels like stdio or HTTP used to carry MCP protocol messages between client and server

Why it matters: MCP creates a portable, interoperable way to plug tools and data into different AI clients without rewriting integrations for each one. By clearly defining what a server can do and how it is invoked, it helps enforce safer capability boundaries and reduces the risk of uncontrolled access to systems or data. It also makes integration and testing easier, since the same MCP server can be exercised by multiple hosts, in automated tests, and in different environments using a consistent protocol.

Connect Measured’s MCP server

Ready to plug Measured into your AI client? See Connect Measured’s MCP server to AI clients for setup instructions for Claude.ai, ChatGPT, Microsoft Copilot, and Gemini Enterprise.