Model Context Protocol (MCP): a plain-English definition
2026-11-10
Definition
The Model Context Protocol, or MCP, is an open standard (see the Model Context Protocol site) that gives AI assistants a common way to connect to outside tools and data sources such as files, calendars, databases and business apps.
Why it exists
Before a shared standard, every AI product needed its own custom connection to every app. MCP defines one format, so a tool built once as an MCP server can work with any assistant that supports the standard. It is similar in spirit to an API, but designed for AI models that choose which tool to call and when. This is a key building block for an AI agent.
Treat each connection as a permission decision. An assistant connected to your email or files can read what that connection allows. Review what each server can see and do, use accounts with limited access and set rules through AI governance. Support varies by assistant, so check current documentation. The answer on whether the Privacy Act applies to AI tools is a good companion.
A hypothetical example
Imagine a financial planning firm that connects an AI assistant to its client file store through MCP. A planner asks for the last three meeting notes for one client, and the assistant fetches only those files.
We design and build connected assistants under AI agents.