MCP, short for Model Context Protocol, is a shared standard that lets AI assistants connect to other apps, files and tools. It is how a chatbot can read your calendar or work inside another program, instead of only chatting.
Think of a universal travel adapter. Without one, every country’s socket needs its own plug. With one, anything plugs in anywhere. Before MCP, every AI app needed a custom connection to every other app. MCP is one agreed plug shape, so any assistant that speaks it can connect to any app that offers it.
In a sentence: “Once I connected the calendar app over MCP, my assistant could find a free slot and book the meeting itself.”
What the term means
A “protocol” is just an agreed set of rules for how two programs talk. MCP is that set of rules for AI. The official MCP site defines it in one line and uses its own plug comparison: “MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems.” and “Think of MCP like a USB-C port for AI applications.”1
Anthropic, the company behind Claude, released it as an open standard in November 2024.2 “Open” means anyone can use it. The MCP site lists both Claude and ChatGPT among the assistants that support it.
A simple example
An app offers an “MCP server”; your assistant plugs into it. A server here is just the piece of software on the app’s side that says what the AI may do: read events, add a note, search files. The MCP site gives an everyday example: “Agents can access your Google Calendar and Notion, acting as a more personalized AI assistant.”1
So you might ask, “Find a free hour on Thursday and add ‘dentist’ to my calendar.” Through MCP, the assistant can check the calendar and add the entry itself. Many apps call these connections “connectors” or “integrations”.
Why it matters to you
Plugging an app in hands the AI real keys. Once connected, the assistant can act, not just talk. The permissions you grant matter more than how carefully you word your request: a polite prompt cannot stop something the connection allows. Skills, repos and MCP explains that point.
Only connect servers you trust. Some MCP servers run as programs on your own computer. The MCP project’s own security guide warns about servers from unknown sources.3 Connections also widen the door for prompt injection, where hidden text tries to steer the AI.
Where the word comes from
The name comes from Anthropic, which introduced the Model Context Protocol on 25 November 2024. Its launch post named the problem MCP was meant to fix: “Every new data source requires its own custom implementation, making truly connected systems difficult to scale.”4 “Context” here means the information an AI is given to work with.
On 9 December 2025 Anthropic handed MCP over to the Agentic AI Foundation, which it describes as “a directed fund under the Linux Foundation, co-founded by Anthropic, Block and OpenAI”. So MCP is no longer run by one company alone.5
Often confused with
| Term | How it differs |
|---|---|
| Skill | Differs: a skill is written know-how the AI reads (a recipe card). MCP is the connection to another app (the socket). See what AI skills are. |
| Agent | Differs: an agent is an AI that takes steps on its own. MCP is one of the ways an agent reaches the apps it works in. |
| Plugin or app | Similar idea: a plugin usually works with one product only. MCP is a shared standard, so one connection can work with many assistants. |
Questions people ask
Is MCP only for Claude?
No. Anthropic started it, but the official MCP site says “AI assistants like Claude and ChatGPT, development tools like Visual Studio Code, Cursor, MCPJam, and many others all support MCP”.6 One connection can therefore work with many assistants.
What is an MCP server?
It is the piece of software on an app’s side that the assistant plugs into. MCP’s own documentation calls it “A program that provides context to MCP clients” and says “MCP servers can execute locally or remotely”: on your own computer, or on the app maker’s computers over the internet.7
Is MCP safe to use?
It is as safe as the servers you connect and the permissions you give them. The MCP project warns that servers “from untrusted sources introduce several critical security risks”, so install only ones from makers you trust.3
Learn more on this site
- Skills, repos and MCP — three words you keep meeting, for people who do not write code.
- AI agents for beginners — what agents can do with connections like MCP, and how to stay safe.
- Skills are a supply chain — why what you install matters.
- The AI glossary and AI for beginners — the basics, in plain words.
QUICK CHECK · 3 QUESTIONS
Pick an answer to see if you got it. Nothing is recorded.
1. What is MCP?
- A shared standard that lets AI assistants connect to other apps, files and tools
- A new chatbot made by one company
- A type of computer chip
Show the answer
Answer: A shared standard that lets AI assistants connect to other apps, files and tools. MCP is an agreed set of rules, like a universal travel adapter, so any assistant that speaks it can plug into any app that offers it.
2. What is an MCP server?
- A big data centre that trains models
- The software on an app’s side that says what the AI may do
- A person who checks every AI answer
Show the answer
Answer: The software on an app’s side that says what the AI may do. The app offers an MCP server, and your assistant plugs into it; the server sets what the AI may read or change.
3. What matters most once you connect an app over MCP?
- How politely you word your request
- How long your prompt is
- The permissions you grant, and trusting the server you connect
Show the answer
Answer: The permissions you grant, and trusting the server you connect. A polite prompt cannot stop something the connection allows, so only connect servers you trust.
Related terms: Eval · Function calling · Human-in-the-loop · Multi-agent system · Orchestrator · Plugin
Related on this site: Before You Connect an AI Agent or Plugin · Check a Skill Before Installing · Skills vs MCP vs Plugins · What Is Function Calling in AI? · What Is a Plugin in AI? · Blender MCP: How to Use Blender With Claude and Other AI
- 1Model Context Protocol, What is the Model Context Protocol (MCP)?, read at source 1 Oct 2026“MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems.” and “Think of MCP like a USB-C port for AI applications.” ↑a“Agents can access your Google Calendar and Notion, acting as a more personalized AI assistant.” ↑b
- 2Anthropic, Introducing the Model Context Protocol, 25 Nov 2024, read at source 1 Oct 2026“Today, we’re open-sourcing the Model Context Protocol (MCP), a new standard for connecting AI assistants to the systems where data lives”. ↑
- 3Model Context Protocol, Security Best Practices, read at source 1 Oct 2026“Local MCP servers with inadequate restrictions or from untrusted sources introduce several critical security risks”. ↑a“Local MCP servers with inadequate restrictions or from untrusted sources introduce several critical security risks”. ↑b
- 4Anthropic, Introducing the Model Context Protocol, 25 Nov 2024, read at source 12 Oct 2026“Every new data source requires its own custom implementation, making truly connected systems difficult to scale.” ↑
- 5Anthropic, Donating the Model Context Protocol and establishing the Agentic AI Foundation, 9 Dec 2025, read at source 12 Oct 2026“Today, we’re donating the Model Context Protocol (MCP) to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation, co-founded by Anthropic, Block and OpenAI.” ↑
- 6Model Context Protocol, What is the Model Context Protocol (MCP)?, read at source 12 Oct 2026“AI assistants like Claude and ChatGPT, development tools like Visual Studio Code, Cursor, MCPJam, and many others all support MCP”. ↑
- 7Model Context Protocol, Architecture overview, read at source 12 Oct 2026“MCP Server: A program that provides context to MCP clients” and “MCP servers can execute locally or remotely.” ↑
The comparisons and the example on this page (the travel adapter, the recipe card and socket, the dentist booking) are this site’s own ways of explaining, not claims from these sources.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK