In Claude Code, Claude Code plugins are packages that bundle several kinds of add-ons into one installable unit. A single plugin can carry any mix of skills, slash commands, hooks, connectors to outside tools (called MCP servers), and subagents. Instead of turning each of those on one at a time, you install the plugin and the whole toolkit becomes available at once. Think of it as a boxed set of capabilities rather than a pile of loose parts.
What Claude Code plugins are
Every capability in Claude Code can be added on its own. A skill teaches Claude a repeatable task. A slash command is a shortcut you type. A hook runs a step automatically at a set moment. An MCP server connects Claude to an outside tool like your CRM or calendar. A subagent is a helper Claude can hand a job to. Setting each of these up individually works fine when it's just you tinkering. It gets slow and error-prone when you want a team of people to all end up with the same setup.
A plugin solves that. It gathers a chosen set of those add-ons into one package with a name and a version. Anyone who installs it gets the exact same collection, configured the same way.
One plugin can carry any combination of these. Install it and they all switch on together.
How you install and manage them
Plugins are handled through the /plugin command. From there you can browse what's available, install a plugin, and turn plugins on or off. When you install one, every piece inside it becomes available in that moment. There's no separate step to wire up each skill or connector by hand.
Open the plugin manager to browse, install, enable, or disable. Click to copy.
The typical actions are worth knowing in plain terms:
- Install adds a plugin and everything it carries.
- Enable and disable turn an installed plugin on or off without removing it, so you can flip capabilities on for one project and off for another.
- List shows what you already have.
Where plugins come from: marketplaces
Plugins are distributed through marketplaces. A marketplace is usually just a git repository, a shared code folder, that lists the plugins available to install from it. You point Claude Code at a marketplace, and then you can install any plugin it offers. Anthropic runs public marketplaces, and a team can host its own -- including a private one that only the company can reach.
A real example
Say your company has settled on a way it wants work done in Claude: a skill that drafts client status reports in your format, a command that files a task into your project tool, a hook that keeps a running log of what changed, and a connection to your CRM. Setting that up on one laptop is an afternoon. Setting it up the same way on ten laptops is a week of screen-shares and a guarantee that two of them end up slightly different.
Instead, you package all of it into one plugin, publish it to your company's marketplace, and tell the team to run /plugin and install it. Now every teammate has the same skills, the same commands, and the same connections, configured identically. When you improve the plugin later, you publish a new version and everyone updates to it. The plugin becomes the way your company's standards travel.
Plugin, skill, or connector: which word means what
These terms get mixed up, so here's the short version. A skill, a command, a hook, and an MCP server are individual capabilities. A plugin is the box that holds a chosen set of them so they install and travel together. If you only ever work solo on one project, you may never need a plugin. The moment you want other people to have the exact setup you built, a plugin is the answer.
