An open standard for plugging AI apps into tools and data in one shared way.
MCP (Model Context Protocol) is a shared set of rules for connecting AI apps to outside tools and data. Anthropic released it as an open standard in November 2024. A protocol is simply an agreed way for programs to talk to each other.
Before, every AI app needed its own custom link to every service. With MCP, a service builds one MCP server, and any app that supports MCP can use it.
OpenAI adopted MCP in March 2025, and in December 2025 it was donated to a new foundation under the Linux Foundation. Easier connections bring a risk, though: plug in an untrustworthy server and the AI may follow bad instructions.
Think of it like this
Phones used to have different charger shapes, so borrowing a friend's charger often didn't work. Now one USB-C cable fits most devices. MCP is like USB-C for connecting AI apps to tools.
For example
In an app that supports MCP, you could connect your calendar and ask, "Sort out my plans for tomorrow."
Common mix-ups
Mix-upMCP is a kind of AI model.
ActuallyMCP itself doesn't think at all. It only sets the format for a tool to announce "here's what I can do" and to exchange requests, so an app can attach the same MCP server whichever model it uses.
Mix-upAnything connected through MCP is safe to use.
ActuallyAnyone can build and publish an MCP server, and by December 2025 there were already more than 10,000 public ones. Before connecting one, check whether you can trust whoever made it and that it gets only the permissions the job really needs.
Try it in class
For teachers at the start of a lesson — or for students working on their own.
- Question
- When a shared standard appears, who saves the most work?
- Activity
- Draw the charger comparison. First connect three AI apps to four services with a line for every pair and count the connectors. Then put an MCP box in the middle and count again. Finally read the November 2024 (E0467), March 2025 (E0549) and December 2025 (E0787) records in order and note who came to share the standard.
Records about this
Source-checked moments from AI history.
The founding projects were MCP, goose and AGENTS.md, at a point when public MCP servers exceeded 10,000
A competitor accepting its rival's standard made it the de facto industry standard
Replaced the combinatorial explosion of N models times M data sources with a single open standard