Claude Plugins Explained: What They Are and How to Use Them

A Claude plugin is a package of Skills, commands, agents and connectors. Here is how it differs from a Skill, how to install one, and how to check it is safe.

You keep seeing the word "plugin" next to Claude, and it is not clear what it means. Is it the same as a Skill? Is it an app? Do you need one? This guide explains Claude plugins in plain English: what is inside one, how it differs from a Skill or a connector, how to install one, and how to check that it is safe before you do.

I checked Anthropic's Claude Code plugin documentation and its plugins guide for claude.ai and Cowork on 11 October 2026. Menus and features move quickly, so I link to the official pages for anything that can change.

What is a Claude plugin?

A plugin is a package. It bundles several useful parts, such as Skills, commands, agents and connectors, so you can add them to Claude in one step. Anthropic's documentation describes a Claude Code plugin as a directory of skills, agents, hooks and MCP servers that installs and loads as one unit.

Think of it like a toolbox instead of a single tool. A Skill is one instruction sheet. A plugin can hold several instruction sheets, plus the connections they need to work.

Why does that matter? Because someone else did the setup. You install one package and get a ready-made working setup, instead of building each part yourself.

What can be inside a plugin?

Anthropic lists these parts. Not every plugin has all of them.

Anthropic's guide says Claude chat, Cowork and Claude Code each load the parts they support and skip the rest. So one plugin can do more in Cowork than in a normal chat. If you are not sure which tool is which, see my guide to Claude Cowork vs Claude Code.

Plugin vs Skill vs connector: what is the difference?

This is the question most people have. Here is the short answer:

• A Skill is a set of instructions for one job. Claude uses it when your task matches.
• A connector links Claude to an outside service, so Claude can read or use it.
• A plugin is the package that can hold Skills, commands, agents and connectors together.

You do not need a plugin to use a Skill. Anthropic's documentation says skills, agents, hooks and MCP servers all work on their own. A plugin is useful when you want several of them packaged as one unit, or when you want to share your setup with a team.

How to add a plugin in Claude (web, desktop and Cowork)

According to Anthropic's guide, you start from the Plugins page:

1. Open Claude or the Claude desktop app and click Customize in the left sidebar.
2. Open the Plugins tab, then Discover, to see the plugins available to you.
3. Select a plugin and read what it contains.
4. Click Add.

Once added, the plugin is saved to your account. Anthropic says its skills and commands are then available in chat, in Cowork and in Claude Code sessions signed in with the same account. To use one, describe the task, or type / in the message box and pick the plugin's skill from the list.

A bundled connector is not switched on automatically. Anthropic notes that adding a plugin does not sign you in to anything. Open the plugin's Connectors tab to connect each one yourself.

To turn a plugin off, use its Disable plugin toggle. To remove it, open its menu and choose Remove. For which plugins appear on which plan, see whether Claude AI is free and Anthropic's pricing page, because the directory's plugins are listed for paid plans.

If you use Cowork day to day, my guide to what Claude Cowork is is a good place to start, and the free Claude Cowork Starter Prompts give you 30 first tasks to try.

How to install a plugin in Claude Code

In Claude Code, you work in the terminal. Anthropic's documentation shows this example, which installs a plugin called commit-commands from its official marketplace:

/plugin install commit-commands@claude-plugins-official

This opens the plugin's details so you can review it and choose a scope first. To browse instead, type /plugin with no name and open the Discover tab.

The scope decides who gets the plugin:

• User scope: you, in every project on this computer.
• Project scope: everyone who works in this project.
• Local scope: you, in this project only.

When it is installed, type / and you will see the plugin's skills in the form /plugin:skill. For this example, that is /commit-commands:commit. There is also a desktop app option and a VS Code option, covered in Anthropic's install guide.

Where plugins come from: marketplaces

A plugin marketplace is a catalog that lists plugins and where to get them. Anthropic's documentation calls it a catalog, not a hosted store. Claude Code adds Anthropic's official marketplace for you the first time you start an interactive session. To add another, you use /plugin marketplace add with its source.

Do not mix this up with the Claude Marketplace website, which is where you browse plugins, connectors and partner products. Anthropic says it is not a marketplace you add with a command. I cover where to find and sell Skills in my guide to the Claude Skills marketplace.

Are Claude plugins safe?

Not automatically, and this is the part to take seriously. Anthropic's Claude Code documentation says a plugin you install can run code on your machine with your user privileges. Its trust warning tells you to make sure you trust a plugin before installing, updating or using it.

Here is what can be inside a plugin, according to Anthropic's plugin security page:

• Hooks run commands on your computer at certain moments.
• MCP servers can start a program on your machine.
• Skills, commands and agents enter Claude's instructions and influence what it does.
• Updates can change files you already reviewed, if auto-update is on.

Anthropic also says its own review covers plugins listed in its directory, but not plugins you add from a marketplace link or upload yourself. Add those only from sources you trust.

Run this check before you install:

1. Check who made it and where the marketplace comes from.
2. Read the "Will install" list in the plugin's details. It names the commands, agents, skills, hooks and servers it adds.
3. Open the plugin's source and read any hooks file and server settings, because the list shows a hook exists but not what it runs.
4. Check what it connects to. Does it need access to your files or accounts? Only connect what the job needs.
5. Try it on harmless data first.

If you decide you no longer trust one, run claude plugin uninstall <plugin> in your terminal, or remove it from Customize in Claude.

Do plugins slow Claude down or use more of your plan?

They can. Anthropic's documentation explains that for each skill, agent and command Claude can use on its own, the name and description sit in Claude's context on every turn. That counts toward your usage and leaves less room, even when you never use the plugin. Plugins in Anthropic's official marketplace show a Context cost estimate in Claude Code before you install.

So do not install everything. Keep a short list of plugins you use every week and disable the rest. In Claude Code, the Installed tab in /plugin has a Not used recently group that helps you find them.

Do you need a plugin?

Here is how I would decide:

• You do one repeating job: a single Skill is enough.
• You want a whole ready-made setup: a plugin may save you hours.
• You want Claude to use an outside tool: you need a connector, which may come inside a plugin.
• You want to share your setup with a team: package it as a plugin.

My honest advice for beginners: start with Skills and one or two plugins from Anthropic's own directory. Learn what they do, then add more only when a real task needs them.

Your next step

Open Customize in Claude, go to Plugins and Discover, and read the details of one plugin without installing it. Find its skills, commands and connectors. Then use the free checklist below to decide whether it is worth adding and how to check it is safe.

Frequently asked questions

What are Claude plugins?

A Claude plugin is a package that can hold Skills, commands, agents and connectors, so you add a ready-made setup in one step. In Claude Code, a plugin can also hold hooks and MCP servers. Anthropic's documentation describes it as a set of components that loads as one unit.

What is the difference between a Claude plugin and a Skill?

A Skill is a set of instructions for one job. A plugin is a package that can contain Skills along with commands, agents and connectors. You do not need a plugin to use a Skill, because Skills work on their own.

How do I install a Claude plugin?

In Claude on the web or desktop app, open Customize, then Plugins, then Discover, select a plugin and click Add. In Claude Code, type /plugin, open the Discover tab and install one. Read what the plugin adds before you confirm.

Are Claude plugins safe?

Not automatically. Anthropic's Claude Code documentation says a plugin can run code on your machine with your user privileges, and that its own review covers plugins in its directory but not ones you add from a link or upload. Check who made a plugin, read what it adds and test it on harmless data first.

Do Claude plugins cost money?

Anthropic's pages I read do not list a general price for plugins, so check each listing. Plugins from the directory appear on Pro, Max, Team and Enterprise plans, and a connector inside a plugin may need its own account or paid plan with that service.

Do I need a plugin to use Claude Skills?

No. Skills, agents, hooks and MCP servers all work on their own. A plugin is useful when you want several of them packaged together or want to share your setup with a team.

Free resource

Plugin vs Skill vs connector table, 10-minute safety check, install steps for Claude and Claude Code, and a plugin tracker

Download the free Claude Plugins Starter Checklist

×

Log in to continue

Log in to unlock this download. It's free — you can sign up in a minute.

Log in → Sign up