Claude plugins are bundled extensions that shape how Claude works for a particular role, project, or team. Rather than wiring up the same skills, connectors, and instructions by hand every time you open a new conversation, a plugin carries all of that in one installable package you switch on once. Maybe you’ve hit the word while poking around the Claude app or Claude Code and weren’t sure what a plugin actually holds — or how it’s any different from a skill or an MCP server. This guide covers what Claude plugins are, what goes inside them, where they run, and how to install and use one, on both the everyday Claude app and Claude Code.
Last updated: July 16, 2026.
What a Claude plugin is, in plain English
The simplest way to picture a plugin is as a container. It doesn’t do one specific thing the way a single command does; it’s a bundle that ships several capabilities together, already configured. Anthropic puts it plainly: a plugin bundles skills, connectors, and sub-agents into a single package, so you get a ready-to-go setup from your first message instead of assembling each piece yourself.
If you’ve ever used a browser extension or a VS Code extension, you already know the shape of this. The extension isn’t the feature — it’s the wrapper that drops a set of features in with one click. A Claude plugin does the same job for Claude. One install brings in a coherent set of tools built around a task (sales, finance, legal, engineering, code review, security scanning) instead of a loose pile of settings you stitch together yourself.
What’s inside a plugin
A single plugin can carry any mix of these parts:
- Skills — Markdown instruction files (
SKILL.md) that teach Claude a method or a set of guidelines, like how your team structures code or writes tickets. Claude pulls a skill in when the task matches it. - Connectors — integrations that hook Claude up to outside services such as Google Drive, Gmail, Slack, or DocuSign, so the right services are already wired in for the workflow.
- Sub-agents — specialized helper agents the main Claude can hand a focused job to, like architectural planning or a security review.
- Hooks — scripts that fire automatically on events, such as when a session starts or right before a tool runs.
- Commands — slash-command shortcuts (say,
/review) that kick off a set workflow.
In Claude Code specifically, plugins can also carry MCP servers, LSP servers for code intelligence, and background monitors. One nuance worth knowing in the consumer apps: skills from a plugin work everywhere, but hooks and sub-agents run only in Cowork — in ordinary chat they show up grayed out.
Plugins vs skills vs MCP vs connectors
This is the distinction the whole ecosystem trips over, and it lands right in Google’s “People also ask” box. Here’s the short version: a plugin is the package, while skills, MCP servers, connectors, hooks, and sub-agents are things a package can contain. You can run a skill or an MCP server on its own without any plugin at all — the plugin just distributes several of them together.
| Term | What it is | Where it runs | Reach for it when |
|---|---|---|---|
| Plugin | A bundle that packages skills, connectors, sub-agents, hooks, and/or MCP servers | Claude app (chat, Desktop, Cowork) and Claude Code | You want a whole role or workflow set up in one install |
| Skill | A Markdown instruction file that teaches Claude a method or standard | Everywhere Claude runs | You want Claude to follow a specific procedure or house style |
| MCP server | A service that gives Claude live access to external tools and data | Claude Code, and Cowork via Anthropic’s cloud | You need Claude to reach a database, an API, or your own tool |
| Connector | A built-in integration to a common service (Gmail, Slack, Drive) | Claude app and Cowork | You want Claude reading or writing a mainstream app |
| Hook / sub-agent | An event script / a delegated helper agent | Cowork (and Claude Code) | You want automation on events, or a focused second agent |
If you only keep one line: the plugin is the box; the skills, servers, and connectors are what’s in it. For a closer look at the pieces, see our explainers on Claude Skills and what an MCP server is.
Why use plugins
The point is setup you don’t have to redo. Without plugins, you re-explain your conventions, reconnect the same services, and re-paste the same instructions at the top of every project. A plugin captures that once and makes it portable: share it with teammates, version it, reuse it across projects, and everyone starts from the same configured baseline instead of a blank page. For teams that repeatability is the real draw — one plugin means the whole group’s Claude behaves the same way on the same kind of work.
Where you can use plugins
Plugins are available on all paid Claude plans — Pro, Max, Team, and Enterprise — as of July 2026. In the consumer product you can install and use them in three places: chat on the web, the Chat tab in Claude Desktop, and Claude Cowork. Developers can also use plugins in Claude Code, the command-line tool that runs in your terminal. Claude Code is the context most of the tutorials floating around actually describe, which is why “Claude plugins” and “Claude Code plugins” get thrown around almost interchangeably even though the app supports them too.
How to install a Claude plugin
The steps differ a little between the Claude app and Claude Code.
In the Claude app (web, Desktop, Cowork):
- Open the Customize menu in the left sidebar — it pulls your plugins, skills, and connectors into one place. In Cowork, open the Cowork tab first, then open Customize.
- Open the Plugins tab.
- Click Browse plugins to see what’s available.
- Click Install on the plugin you want.
You can also upload a custom plugin file if you built one yourself or got one from a colleague. On Claude Desktop and in Cowork, plugins you add yourself are saved locally on your machine.
In Claude Code (terminal):
Claude Code works through marketplaces and the /plugin command. Anthropic runs two public ones: claude-plugins-official, a curated set, and the community marketplace, which you add with /plugin marketplace add anthropics/claude-plugins-community and then install from using /plugin install. If you don’t see the /plugin command, update Claude Code to the latest version. You can also add any marketplace hosted in a GitHub repo by pointing at its git URL — the same “add from a repository” option the app offers under Customize → Plugins → + → Add marketplace.
How to use an installed plugin
Once a plugin is installed, its skills line up next to Claude’s built-in ones. In chat and in Cowork, type / or click the + button to see the skills your installed plugins added, then click one to read what it does. Skills can also fire on their own when the task matches their description — that’s the model-invoked behavior, so you don’t always have to call them by name. In Claude Code, plugin skills are namespaced so they don’t collide, which means you trigger them as /plugin-name:skill-name, and /reload-plugins picks up changes without a restart.
Examples of Claude plugins
Anthropic ships a growing library of plugins for common knowledge work — sales, finance, legal, marketing, HR, engineering, design, operations, and data analysis — each pre-loaded with the skills and connectors that fit the function. The Knowledge Work marketplace is added by default in the app, and you can add others like Financial Services or Legal yourself. There’s also Plugin Create, a plugin whose whole job is helping you build one from scratch. The full set of Anthropic-built plugins lives in the open anthropics/knowledge-work-plugins repository on GitHub.
On the Claude Code side, community directories have grown fast. The official anthropics/claude-code repo documents the plugin system, and third-party hubs now index hundreds of community plugins and marketplaces. The developer picks that come up most are practical tooling: a TypeScript language-server plugin, security-guidance plugins, documentation helpers. We’ll cover the best of these in a dedicated roundup; for now, the safest sources are the two Anthropic marketplaces and repos you can actually read before installing.
What a plugin looks like under the hood
If you’re curious what you’re actually installing, a Claude Code plugin is just a folder with a manifest. The minimum is a .claude-plugin/plugin.json file plus whatever components you’re shipping:
my-first-plugin/
├── .claude-plugin/
│ └── plugin.json <- the manifest (name, description, version)
└── skills/
└── hello/
└── SKILL.md <- one skill: frontmatter + instructions
The manifest itself is small:
{
"name": "my-first-plugin",
"description": "A greeting plugin to learn the basics",
"version": "1.0.0",
"author": { "name": "Your Name" }
}
Only name is required — it doubles as the namespace for the plugin’s skills. Everything else (agents/, hooks/, .mcp.json, commands/) sits at the plugin root, never inside .claude-plugin/. That’s the whole trick: a plugin is a directory, a small JSON file that names it, and the components you drop in beside it. You can test one locally with claude --plugin-dir ./my-first-plugin before you ever publish it. The full specs are in Anthropic’s Plugins reference.
Are Claude plugins safe to install?
Mostly, with one caveat that’s worth taking seriously: plugins can bundle local MCP servers that run on your computer with the same permissions as any other program you run. That’s access you want to grant on purpose. Install plugins from sources you trust, lean on the Anthropic marketplaces and repos you can inspect, and be wary of a random marketplace you can’t vet. On Enterprise plans, admins can limit which plugins are installable or switch off local MCP servers entirely. The folder-and-manifest structure above is also what makes vetting realistic — you can open a plugin and read exactly what it ships before you enable it.
FAQ
What’s the difference between Claude skills and plugins?
A skill is a single Markdown instruction file that teaches Claude one method or standard. A plugin is a package that can hold many skills, plus connectors, sub-agents, hooks, and MCP servers. You can use a skill without a plugin; the plugin exists to distribute a set of them together, versioned and shareable.
How do I install a Claude plugin?
In the Claude app, open Customize in the left sidebar, go to the Plugins tab, click Browse plugins, and hit Install. In Claude Code, add a marketplace — for example /plugin marketplace add anthropics/claude-plugins-community — and run /plugin install.
Are Claude plugins free?
Plugins are a feature of the paid Claude plans (Pro, Max, Team, Enterprise) as of July 2026, and the Anthropic-built ones are free to install. Community plugins set their own terms, so check each one’s license and source.
Where can I find Claude plugins on GitHub?
Anthropic’s knowledge-work plugins are at github.com/anthropics/knowledge-work-plugins, the Claude Code plugin system is documented in github.com/anthropics/claude-code, and community submissions land in the anthropics/claude-plugins-community catalog.
Do I need Claude Code to use plugins?
No. Plugins work in the regular Claude app — web chat, the Chat tab in Claude Desktop, and Cowork — not just in the terminal. Claude Code adds developer-focused pieces like MCP and LSP servers, but the install-and-use flow lives in the app too.
Wrapping up
Claude plugins are the packaging layer for everything else Claude can be extended with: skills, connectors, sub-agents, hooks, and MCP servers, bundled so a whole workflow installs in one step. Pick an Anthropic marketplace plugin that matches your work, install it from Customize → Plugins, and trigger its skills with /. When you want to understand the pieces a plugin bundles, our Claude Skills explainer and MCP server guide go a level deeper, and if you’re building agentic workflows, our AI agents explainer ties it all together.