Cursor Plugin とは何か — 入れたものを試して分かったこととおすすめThe term "plugin" in Cursor conflates VS Code extensions, MCP servers, and…

Ask a colleague whether they have "installed a Cursor plugin," and the answers tend to diverge in a way that reveals a genuine gap in shared vocabulary. One person hears the question and thinks of the ESLint extension; another assumes you mean adding a single MCP server; a third pictures downloading something from a Marketplace. The word "plugin" is identical, but the thing it points to is not, and that ambiguity is becoming more common as Cursor's ecosystem grows.

This article, published on Zenn, starts from that terminological mismatch and works toward a clearer picture of what a Cursor Marketplace Plugin actually is—specifically, what bundle of components it represents—before reporting on hands-on testing of roughly four or five such Plugins and offering recommendations.

The core problem is that three distinct layers now share one casual label. The first is the VS Code Extension. Cursor is built on the same foundation as Visual Studio Code, so it inherits the familiar extension model, and tools like ESLint, Prettier, or language-specific packages install in much the way developers already expect. The second layer is the Model Context Protocol server, usually shortened to MCP. Introduced by Anthropic and since adopted across several AI tools, MCP is an open standard that lets an AI assistant connect to external data sources, APIs, and actions through a consistent interface. Adding "an MCP" means wiring up one of these servers so the model can, for example, query a database or call a service. The third layer is Cursor's own Plugin, distributed through the Cursor Marketplace as a package.

The article's central clarification is that a Marketplace Plugin is best understood not as a single feature but as a bundle. Rather than being only an extension or only an MCP entry, a Plugin appears to package several of these elements together—potentially combining editor configuration, rules or instructions for the AI, and MCP connections—so that installing one item sets up a coordinated group of capabilities at once. Framing the Plugin as "a bundle of what?" is the lens the piece uses to cut through the confusion.

That distinction matters for practical reasons. If a user believes a Plugin is just another VS Code extension, they may misjudge what it changes about the assistant's behavior or what external access it introduces. Conversely, treating everything as "MCP" obscures the editor-side pieces that a Plugin can carry. Knowing which layer you are touching helps with troubleshooting, security review, and deciding whether a given addition is worth keeping.

Some background helps situate this. Cursor is an AI-centric code editor that forks the VS Code base, which is why extension compatibility is a recurring theme; many VS Code extensions are made available through the Open VSX registry rather than Microsoft's own marketplace. MCP, meanwhile, has become a widely referenced pattern for connecting language models to tools, and its growing adoption is part of why "just add an MCP" has entered everyday developer speech. A dedicated Marketplace that packages these pieces is a logical next step, and it mirrors a broader industry pattern of turning loose configuration into installable, shareable units.

From the hands-on portion, the author reports installing on the order of four to five Marketplace Plugins and drawing conclusions from that experience. The framing suggests the value of a Plugin depends heavily on what it bundles and how well those parts work together in practice, rather than on the label alone. Readers looking for concrete recommendations are pointed toward the results of that testing.

For anyone adopting Cursor in a team setting, the takeaway is less about any single Plugin and more about establishing shared language. Agreeing on whether "plugin" means a VS Code extension, an MCP server, or a Marketplace Plugin can prevent miscommunication when someone asks what has been installed. As these three layers continue to coexist, being precise about which one is in play is likely to remain a useful habit, and the Marketplace's bundling approach may make that precision even more important over time.