Plugins

Plugins for coding agents

Every plugin we list, with the install command for each agent and which agents it works with.

15 listed

  • Expo plugin

    The Expo plugin is a collection of agent skills for working with Expo projects and Expo Application Services.

    • Claude Code
    • Plugin

    expo · 3,880 stars

  • Exa plugin

    The Exa plugin brings Exa's web search and crawling to Claude Code through its MCP server.

    • Claude Code
    • Plugin

    Exa · 5,204 stars

  • Atomic Agents plugin

    Atomic Agents is a framework for building AI agents from small, composable parts, and this is its Claude Code plugin, with the repository tagline Building AI agents, atomically.

    • Claude Code
    • Plugin

    Eigenwise · 6,314 stars

  • Huggingface Skills plugin

    Hugging Face Skills is the plugin form of the huggingface/skills repository, described as a way to give agents the power of the Hugging Face ecosystem.

    • Claude Code
    • Plugin

    huggingface · 11,240 stars

  • Fullstack Dev Skills

    Fullstack Dev Skills is a plugin bundling 67 specialised skills for full-stack developers, pitched as a way to turn Claude Code into a pair programmer.

    • Claude Code
    • Plugin

    jeffallan · 12,143 stars

  • ccstatusline

    ccstatusline is a customisable status line for the Claude Code command line, with themes and powerline styling.

    • Claude Code
    • Plugin

    sirmalloc · 13,475 stars

  • Compound Engineering Plugin

    The Compound Engineering plugin is described as the official plugin from Every for Claude Code, Codex, Cursor and more.

    • Claude Code
    • Plugin

    EveryInc · 25,789 stars

  • Claude HUD

    Claude HUD is a Claude Code plugin that shows what the session is doing, including context usage, active tools, running agents and todo progress.

    • Claude Code
    • Plugin

    Jarrod Watts · 29,035 stars

  • Official Plugin Directory

    The official plugin directory is an Anthropic-managed repository of Claude Code plugins, described as a directory of high quality plugins.

    • Claude Code
    • Plugin

    Anthropic · 45,440 stars

  • Chrome Devtools Mcp plugin

    Chrome DevTools for coding agents is the project description, and this entry is the Claude Code plugin form of the ChromeDevTools MCP server.

    • Claude Code
    • Plugin

    ChromeDevTools · 54,503 stars

  • Superpowers plugin

    This is the marketplace listing for Superpowers, the same project as our skill entry but reached through Claude Code's plugin system.

    • Claude Code
    • Plugin

    obra · 309,408 stars

  • Desktop Commander plugin

    This is the Claude Code plugin packaging of Desktop Commander, listed in the claude-plugins-official marketplace and installed with /plugin install desktop-commander@claude-plugins-official.

    • Claude Code
    • Plugin

    Desktop Commander · 10,174 stars

  • Deepeval plugin

    DeepEval is described as the LLM evaluation framework, and this entry is its Claude Code plugin from Confident AI, listed in the claude-plugins-official marketplace and installed with /plugin install deepeval@claude-plugins-official.

    • Claude Code
    • Plugin

    Confident AI · 19,203 stars

  • Hyperframes plugin

    Hyperframes, from HeyGen, is a Claude Code plugin whose repository describes the idea as write HTML and render video, built for agents.

    • Claude Code
    • Plugin

    HeyGen · 76,031 stars

  • Mattpocock Skills plugin

    Skills, from Matt Pocock, is a Claude Code plugin packaging the skills from his own agents directory, and the GitHub description calls them skills for real engineers.

    • Claude Code
    • Plugin

    Matt Pocock · 307,443 stars

A Claude Code plugin is a directory of skills, subagents, hooks, MCP servers and other parts that Claude Code installs as one unit, usually from a marketplace. Run /plugin and open the Discover tab, or install by name with /plugin install <name>@<marketplace>. A plugin runs code with your permissions, so review it first. We checked Anthropic's plugin docs and repos on 31 October 2026.

Quick facts

What it bundles

Skills, agents, hooks, MCP servers, LSP servers, a hooks module (a "mod"), monitors, a bin/ directory

Manifest

.claude-plugin/plugin.json

Browse

/plugin, Discover tab; or claude.com/marketplace/plugins on the web

Install in a session

/plugin install <name>@<marketplace>

Install from a shell

claude plugin install <name>@<marketplace> [--scope project|local]

Manage from a shell

claude plugin enable, disable, uninstall

Reload without restarting

/reload-plugins

Scopes

User (default), project, local

Official marketplace

claude-plugins-official (repo anthropics/claude-plugins-official, Apache-2.0)

Community marketplace

claude-community (repo anthropics/claude-plugins-community), added by hand

Cost

Free to install; each plugin adds context and usage

What a plugin is

Skills, subagents, hooks and MCP servers each work on their own. A skill saved in ~/.claude/skills/ is already available in every project. A plugin is the way to package several of them together, so one command installs a whole setup, and updates arrive from the marketplace.

Anthropic's overview names the parts a plugin can hold:

Part

What it does

Skills

SKILL.md instructions Claude loads when relevant; also run as commands

Agents

Subagent definitions Claude can delegate to

Hooks

Commands that run at points in Claude Code's lifecycle, such as after every edit

A hooks module (mod)

Hooks written as JavaScript functions, which can also draw panes and add commands

MCP servers

Tool servers Claude connects to while the plugin is enabled

LSP servers

Language servers started by the plugin, so Claude can see type errors after edits

A plugin skill appears with the plugin name as a prefix. A skill review in a plugin my-plugin runs as /my-plugin:review.

The cost of an installed plugin

This is the part people overlook. An enabled plugin is part of every session, not only the ones where you use it:

  • The name and description of each skill, agent and command that Claude can invoke on its own sits in context on every turn. Those tokens count toward your usage.
  • MCP servers the plugin defines run alongside each session where it is enabled.
  • Hooks fire at their events.
  • Whatever the plugin runs, it runs as you.

Check the footprint before and after. Plugins in the official marketplace show a context cost estimate when you open them from the Marketplaces tab in /plugin. The Installed tab has a Not used recently group, and /skill-doctor shows what each skill costs. To stop a plugin without uninstalling it, run claude plugin disable <name>.

Where plugins come from: marketplaces

A marketplace is a catalog defined by a .claude-plugin/marketplace.json file in a repository. Anthropic's docs describe these tiers:

Marketplace

Name to type after @

What is in it

How you get it

Official

claude-plugins-official

Plugins Anthropic maintains, plus plugins from partners and other authors

Claude Code adds it the first time you start an interactive session, unless policy blocks it

Community

claude-community

Third-party plugins authors submitted to Anthropic

Add it with /plugin marketplace add anthropics/claude-plugins-community

Topic marketplaces

Varies

For example anthropics/skills and anthropics/knowledge-work-plugins

/plugin marketplace add anthropics/skills

Third-party

The name in its marketplace.json

Plugins from their authors' own repos

/plugin marketplace add <owner>/<repo>

Anthropic notes that most of the official marketplace comes from partners and other authors: tool vendors publish plugins that connect Claude Code to their services, and Anthropic maintains a smaller set of its own, with examples such as commit-commands, code-review and feature-dev, plus language server plugins. Anthropic does not review third-party marketplaces.

The names claude-plugins-official and claude-community are accepted only for marketplaces sourced from github.com/anthropics/ repositories, so a third-party marketplace cannot pose as an Anthropic one. A marketplace's name tells you who publishes the catalog, not what each plugin does.

A note on an old one. The claude-code-plugins marketplace in anthropics/claude-code is deprecated. If a tutorial tells you to run /plugin marketplace add anthropics/claude-code, skip that step and use claude plugin install <plugin>@claude-plugins-official. The plugin once called ralph-wiggum is now ralph-loop, and claude-opus-4-5-migration is retired.

Install a plugin, step by step

In a session

  1. Run /plugin.
  2. Open Discover and search.
  3. Open a plugin, read what it adds, choose a scope and confirm.
  4. Run /reload-plugins if it does not appear.

Or by name: /plugin install <name>. That looks the name up in your marketplaces and opens the details; nothing installs until you choose a scope and confirm.

From your shell

claude plugin install formatter@your-org --scope project

On success it prints Successfully installed plugin: formatter@your-org (scope: project). Plugins installed this way load the next time Claude Code starts, or on /reload-plugins in an open session. Some plugins install by running a command that their marketplace names; Claude Code shows the command and asks you to accept it, which a script cannot do.

On a machine where nobody has opened an interactive session, the official marketplace is not registered yet. A setup script should first run claude plugin marketplace add anthropics/claude-plugins-official.

Add a third-party marketplace

/plugin marketplace add owner/repo
/plugin install some-plugin@marketplace-name

Claude Code prints the marketplace name once it has added it.

Scopes

Scope

Who gets it

Recorded in

User (default)

You, in every project on the machine

enabledPlugins in ~/.claude/settings.json

Project

Everyone in the repository

.claude/settings.json (commit it)

Local

You, in this repository only

.claude/settings.local.json

Committing a project entry turns the plugin on for collaborators but does not download it. Each one runs claude plugin install <name>@<marketplace> --scope project once. Local overrides project, and project overrides user.

The terminal, the desktop app's local sessions and the VS Code extension on one computer read the same settings files, so a user-scope install in one shows up in the others.

Update, disable, remove

claude plugin disable formatter@your-org
claude plugin enable formatter@your-org
claude plugin uninstall formatter@your-org --scope project

Auto-update is per marketplace. When it is on, a plugin you reviewed can change on disk later. For anything sensitive, turn auto-update off for that marketplace and update on purpose.

Plugins worth knowing

We list only what we found in Anthropic's own docs and repos.

Plugin or source

What it is

Notes

anthropics/claude-plugins-official

The official directory

Apache-2.0. Internal plugins in /plugins, partner and community ones in /external_plugins

anthropics/skills

Anthropic's public skills repo as a marketplace

Install with /plugin marketplace add anthropics/skills, then document-skills or example-skills

Slack plugin

Skills plus the Slack MCP server

/plugin install slack@claude-plugins-official

Figma plugin

Skills plus the Figma MCP server

claude plugin install figma@claude-plugins-official

code-review, commit-commands, feature-dev

Plugins Anthropic names as its own

See the slash commands guide for built-in alternatives

Language server plugins

Type errors after edits, navigation by symbol

Anthropic documents a code intelligence page

security-guidance, claude-security

Security review plugins documented by Anthropic

Read their docs pages before use

wshobson/agents

A multi-harness plugin marketplace

MIT; works with Claude Code, Codex, Cursor, OpenCode, Copilot and more

VoltAgent subagents

100+ subagents as installable plugins

MIT

Claude Code templates

A CLI catalog, with its own plugin dashboard

MIT; community

Our plan: browse Discover for the services you already use (Slack, Figma, Sentry, Supabase), install their plugin from the official marketplace, and leave the long tail until you have a need.

Security: read this before you install

Anthropic's plugin security page opens with one line: a plugin you install can execute arbitrary code on your machine with your user privileges. What an installed plugin can do:

  • Hooks run as shell commands around tool calls and other events.
  • Monitors run as background shell commands that Claude Code starts on its own.
  • Mods run JavaScript inside Claude Code with your permissions.
  • MCP and LSP servers start as processes on your machine.
  • A `bin/` directory is added to the PATH of the Bash tool, so Claude's commands can run any executable there.
  • Skills, commands and agents enter Claude's context as instructions and shape what it does with the tools it has.
  • Updates can change the files you reviewed if auto-update is on.

The docs state the important limit: permission rules and the sandbox cover the tool calls Claude makes, not the code a plugin runs by itself. Hooks, monitors and server processes run with your full user permissions and outside the sandbox.

Anthropic's official directory adds its own warning: Anthropic does not control what MCP servers, files or other software are included in plugins and cannot verify that they will work as intended or will not change.

Practical habits:

  1. Read `plugin.json`, `hooks/`, `.mcp.json` and `bin/` before you install. That is where code lives.
  2. Prefer the vendor's plugin for a service you use, from the official marketplace.
  3. Pin or disable auto-update for marketplaces you do not control.
  4. Use project scope deliberately. A committed entry asks every collaborator to install it.
  5. Keep the list short. Every plugin is context, processes and trust.
  6. Remove what you stop using. Check the Not used recently group.
  7. Teams can set policy. Anthropic documents organization controls for allowing the official marketplace and your own.

Do you need a plugin or just a file?

You want

Use

One reusable procedure

A skill in .claude/skills/

One reusable prompt you type

A command

One specialist worker

A subagent file

One service connection

An MCP server

A tested bundle for a team

A plugin

Versioned releases and updates

A plugin in a marketplace

Anthropic's own advice is the same: install a plugin to get a setup someone else built with one command, make one to give your setup to teammates or publish versioned releases.

Build and publish your own

A minimal plugin layout, as shown in Anthropic's official directory README:

plugin-name/
  .claude-plugin/plugin.json   required metadata
  .mcp.json                    optional MCP servers
  commands/                    optional slash commands
  agents/                      optional subagents
  skills/                      optional skills
  README.md

Docs cover creating a plugin from an empty directory, testing it without a marketplace, converting an existing .claude/ setup, adding components, declaring dependencies with version ranges such as ^1.2, writing evals with claude plugin eval, and measuring token cost and usage. To list in Anthropic's directory, third-party partners submit through a form and must meet quality and security standards.

One rule worth knowing: a marketplace entry's name is an immutable slug. Renaming it breaks existing installs with a plugin-not-found error, so use displayName for labels and the top-level renames map if a rename is unavoidable.

Limits and known issues

  • Marketplace not found. On a fresh machine the official marketplace may not be registered. Add it by hand.
  • Updates change what you reviewed. See auto-update above.
  • Context creep. Description text from many plugins adds up on every turn.
  • Quality varies in third-party marketplaces. Anthropic does not review them.
  • Sync across surfaces. Plugins from your claude.ai account sync to Claude Code; cloud sessions load what the repository enables.
  • Namespacing. Plugin skills carry the plugin name as a prefix, which makes commands longer to type.
  • Not the same as the IDE extension. The Claude Code VS Code extension and the JetBrains plugin are separate from Claude Code plugins.

Our take

We read Anthropic's plugin documentation and repos for this page. We did not install and test the plugins named here.

Plugins solve a real problem: distributing a setup. For a team, putting your skills, agents and MCP config in one versioned plugin beats telling everyone to copy five folders. For a solo developer, a plugin is often unnecessary. A skill file in ~/.claude/skills/ does the same job with less trust and less context.

The trust model is the thing to take seriously. Skills are text. Plugins can be programs. The distinction in Anthropic's own docs is clear, and we would follow it: official marketplace for vendors you already trust, review everything else, and keep the number small.

The healthy pattern we would aim for is three plugins you actively use, each from a source you can name.

Frequently asked questions

What is a Claude Code plugin? A directory of skills, agents, hooks, MCP servers or other components installed as one unit.

How do I install a plugin? /plugin then Discover, or /plugin install <name>@<marketplace>.

Are plugins safe? They can run code with your privileges. Review before installing.

Do I need a plugin or just a skill? Only use a plugin to bundle several parts or share with a team.

Official vs community marketplace? Official has Anthropic's plugins plus partners; community holds third-party submissions and is added by hand.

*Product names are used descriptively. We are not affiliated with Anthropic.*

Questions

What is a Claude Code plugin?

A directory of skills, agents, hooks, MCP servers or other components that Claude Code installs and loads as one unit, usually from a marketplace.

How do I install a plugin?

Run /plugin and use the Discover tab, or run /plugin install <name>@<marketplace>, or claude plugin install <name>@<marketplace> in your shell.

Are plugins safe?

A plugin can run code on your machine with your user privileges, through hooks, monitors, MCP servers and its bin directory. Review it before you install, whichever marketplace it comes from.

Do I need a plugin or just a skill?

Skills, subagents, hooks and MCP servers all work on their own. Use a plugin when you want several of them packaged as one unit, shared with a team, or versioned.

What is the difference between the official and community marketplaces?

Official (claude-plugins-official) holds plugins Anthropic maintains plus partner and other authors. Community (claude-community) holds third-party plugins authors submitted to Anthropic, and you add it yourself.

Part of the Wbcom Designs family

The all-in-one WordPress community stack

Also ours: wbcomdesigns.comvapvarun.combrndle.com