Claude Plugins Explained: One Install, Many Powers

| Source: Claude Plugins Documentation

Imagine adding a full set of tools to Claude with one click. That is what a Claude plugin does.

It bundles skills, connectors, commands, and agents into one package, so that you set it up once and use it again and again.

Read ahead to learn how it works.

[The Four Parts]

A plugin has four parts. Skills are instructions Claude follows. Connectors link Claude to outside tools. Commands are named actions you can run. Agents are helpers that take on part of a task.

Together, they work like a toolbox you add to Claude in one step.

[How to Add One]

In claude(.)ai or the desktop app, open the Customize page and then select Plugins.

Use Discover to browse and select Add. You can also add a marketplace, which is a Git repository of plugins, or upload a plugin as a .zip or .plugin file.

[Connect Your Apps]

Here is the step to remember. Adding a plugin does not sign you in to your apps. Open the plugin, go to its Connectors tab, and connect each one yourself.

The skills load anyway, but a skill that needs an app cannot reach it until it shows Connected.

[Use It Your Way]

Using a plugin is simple. In chat, describe your task and Claude loads the right skill, or type / to pick one from the list.

Agents work only in Cowork and Claude Code, not in chat.

You can turn a plugin off or remove it at any time.

[Stay Safe]

Plugins in Anthropic’s directory get a security scan on every version, and a person reviews each new listing.

Plugins you add from a link or upload are not reviewed by Anthropic. Add those only from sources you trust.

13 comments

Claude Plugins Explained: One Install, Many Powers