What is MCP? The Model Context Protocol in plain English

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to the tools and data you already use. Anthropic introduced it in November 2024, and it is now supported by most major AI clients, including Claude, ChatGPT, Cursor and VS Code.

An MCP server is a small program that exposes a set of named tools, such as "search issues" or "create a page", to an AI client. The client shows the AI what each tool does, and the AI decides when to call one on your behalf.

Remote vs local servers

A remote (hosted) server runs on the vendor’s infrastructure. You connect with a URL and usually sign in with OAuth, the same way you would authorize any app. This is the easiest and generally the safest setup.

A local server runs on your own machine, typically started by your AI client with a command like npx or uvx. Local servers often need an API key or personal access token pasted into a config file, and they run with whatever access that key has.

MCP servers vs Claude connectors

In Claude, a connector is how you add an MCP server without editing config files. Connectors listed in Claude’s directory are reviewed by Anthropic, and many are built by the vendor itself. Under the hood, a connector is still an MCP server.

Why "official" matters

Anyone can publish an MCP server for any product. Registries list thousands of them, and many are built by individuals with no link to the vendor. An official server is published by the company that makes the product, so it is maintained alongside the product and its permissions are designed by people who know the API.

That is the question MCP or Not answers for each tool: does the vendor publish a server or connector, what can it actually do, and is it safe to connect? We only call something official when the vendor’s own docs or repository confirm it.