Remote OpenClaw Blog
Claude Plugins: What They Are and How to Install Them
7 min read ·
Claude plugins are shareable packages that extend Claude Code with any combination of skills, slash commands, subagents, hooks, and MCP servers. You install them in two steps: register a catalog with /plugin marketplace add owner/repo, then install individual plugins with /plugin install plugin-name@marketplace-name.
What Are Claude Plugins?
A Claude plugin is a self-contained directory that packages Claude Code extensions so they can be versioned, shared, and installed as one unit. According to Anthropic's launch announcement on October 9, 2025, plugins are "a lightweight way to package and share any combination of" slash commands, subagents, MCP servers, and hooks. Since then the format has grown to include Agent Skills, LSP servers for code intelligence, background monitors, and default settings.
The anatomy is simple. The manifest lives at .claude-plugin/plugin.json and holds the plugin's name, description, and version. Everything else sits at the plugin root: a skills/ directory with SKILL.md folders, an agents/ directory for subagent definitions, hooks/hooks.json for event handlers, and a .mcp.json file for bundled MCP servers, per the official plugins documentation.
One detail trips people up: plugin skills are always namespaced. Install a plugin named commit-commands and its commit skill becomes /commit-commands:commit, not /commit. Namespacing prevents two plugins from colliding when they ship skills with the same name.
Plugins vs Skills vs MCP Servers
Plugins are the packaging layer, skills are instructions, and MCP servers are live tool connections; a plugin can contain both of the other two. If you only need one prompt-level capability, a standalone skill in .claude/skills/ is enough. If you need Claude to talk to an external system, that is an MCP server job. Plugins matter when you want to distribute several of these pieces together with versioned updates.
| Claude Plugins | Skills | MCP Servers | |
|---|---|---|---|
| What it is | A distributable bundle of extensions | Markdown instructions (SKILL.md) Claude loads on demand | A running process that exposes tools and data over the Model Context Protocol |
| Can contain | Skills, slash commands, subagents, hooks, MCP servers, LSP servers | Instructions, scripts, reference files | Tools, resources, prompts |
| Install command | /plugin install name@marketplace | Drop a folder in .claude/skills/ | claude mcp add name -- npx -y package |
| Invocation | Namespaced, e.g. /plugin-name:skill | Direct, e.g. /deploy, or model-invoked | Claude calls tools automatically |
| Best for | Sharing a full workflow with a team or community | Personal or project-specific instructions | Connecting Claude to APIs, databases, and services |
We cover the server side in depth in our Claude Code MCP guide, and the MCP server directory indexes 13,800+ servers if you want to browse what exists.
How to Find and Install Claude Plugins
Installing a Claude plugin is a two-step process: add a marketplace, then install plugins from it. A marketplace is just a git repository or URL with a .claude-plugin/marketplace.json catalog file, so anyone can publish one.
The official Anthropic marketplace, claude-plugins-official, is registered automatically the first time you start Claude Code. For anything else, run these inside a Claude Code session:
# Add a marketplace from GitHub (owner/repo format)
/plugin marketplace add anthropics/claude-plugins-community
# Install a plugin from a marketplace you added
/plugin install plugin-name@claude-community
# Install from the official marketplace (already registered)
/plugin install github@claude-plugins-official
# Activate changes without restarting
/reload-plugins
Running /plugin on its own opens the plugin manager with four tabs: Discover, Installed, Marketplaces, and Errors. Since Claude Code v2.1.145 the details pane shows a "Will install" list of every command, agent, skill, hook, and MCP or LSP server a plugin adds, plus a context cost estimate, so you can review exactly what you are trusting before you confirm. Each install prompts for a scope: user (all your projects), project (shared with collaborators via .claude/settings.json), or local (just you, just this repo).
The hard part is discovery, because plugins are scattered across thousands of GitHub repositories. The Remote OpenClaw plugin directory indexes 3,600+ plugins across 1,400+ marketplaces, is searchable by category and ecosystem, and shows the exact install command for each entry. For plugins that mainly wrap MCP servers, the MCP directory is often the faster path.
Best Claude Plugins to Try First
The most installed Claude plugins as of mid-2026 are frontend-design, Superpowers, and Context7, based on install counts from the official plugin directory compiled by Composio's June 1, 2026 roundup. A few standouts worth knowing:
- frontend-design (829,316 installs): Anthropic's design skill that pushes Claude toward distinctive, production-grade UI instead of generic AI styling.
- Superpowers (752,120 installs): a community plugin by Jesse Vincent that adds structured brainstorming, planning, and debugging skills.
- Context7 (348,660 installs): pulls current, version-specific library documentation into context via the Context7 MCP server.
- code-review and pr-review-toolkit: Anthropic-built review workflows; pr-review-toolkit ships specialized subagents for distinct review concerns.
- commit-commands: git workflows like
/commit-commands:committhat stage, message, and commit in one step. - LSP plugins (typescript-lsp, pyright-lsp, rust-analyzer-lsp, and more): give Claude real code intelligence, including type errors after every edit, if you have the language server binary installed.
- External integrations (github, linear, notion, figma, vercel, supabase, sentry, slack): official plugins that bundle pre-configured MCP servers so you skip manual server setup.
Anthropic also runs a community marketplace at anthropics/claude-plugins-community, where third-party plugins land after automated validation and safety screening, with each plugin pinned to a specific commit SHA. Community catalogs like Dan Avila's and Seth Hobson's, both name-checked in Anthropic's launch post, remain popular sources for agent-heavy setups.
Troubleshooting Claude Plugins
Most Claude plugin problems come down to an outdated CLI, a stale marketplace cache, or a missing binary. Work through these in order:
- /plugin command not recognized: your Claude Code is too old. Update with
npm install -g @anthropic-ai/claude-code@latest(orbrew upgrade claude-code) and restart the terminal. - Plugin not found in any marketplace: refresh the catalog with
/plugin marketplace update marketplace-name, or add it first with/plugin marketplace add owner/repo. - Skills not appearing after install: run
/reload-plugins. If that fails, clear the cache withrm -rf ~/.claude/plugins/cache, restart, and reinstall. - Executable not found in $PATH: LSP plugins need the language server binary installed separately, for example
goplsfor Go orrust-analyzerfor Rust. Check the /plugin Errors tab. - Installed but never invoked: remember the namespace. Run
/helpto see skills listed under/plugin-name:skill-name.
If Claude Code itself is misbehaving rather than a plugin, start with our Claude Code guide and the official discover-plugins documentation, which has a full troubleshooting section.
Limitations and Tradeoffs
Plugins are highly trusted components that execute arbitrary code on your machine with your user privileges, and Anthropic explicitly warns it does not verify what third-party plugins contain. Only install from sources you trust, and review a plugin's homepage and "Will install" list before confirming.
Two other costs are easy to miss. First, every enabled plugin adds tokens to your context window on every turn, which is why newer Claude Code versions surface a context cost estimate and flag plugins "Not used recently." Second, plugins are overkill for personal, single-project tweaks; a plain skill in .claude/ is faster to iterate on and skips the namespace prefix. Package it as a plugin only when you want to share or version it.
Related Guides
- Claude Code MCP: How to Add and Manage MCP Servers
- Claude Code Guide: CLI, Desktop, Web, and Real-World Fit
- Claude Code Best Practices in 2026
- How to Find MCP Servers From Inside Claude Code
Go deeper
The operator playbooks
Production-ready PDF guides for OpenClaw and Hermes Agent — $19.99 each.
Skills for this topic
Browse all skills →Frequently Asked Questions
Are Claude plugins free?
Yes, the plugin system itself is free and built into Claude Code, and the plugins in Anthropic's official and community marketplaces are free to install. Some plugins wrap MCP servers for paid services, so the underlying service (for example Sentry or Figma) may still require its own account or subscription.
What is the difference between Claude plugins and skills?
A skill is a single Markdown instruction set Claude loads on demand, while a plugin is a distribution format that can bundle many skills plus agents, hooks, and MCP servers. Standalone skills live in .claude/skills/ and use short names like /deploy ; plugin skills are namespaced like /my-plugin:deploy .
Are Claude plugins safe?
Plugins run with your user privileges and can execute arbitrary code, and Anthropic does not verify third-party plugin behavior. Prefer the official and community marketplaces, which apply curation or automated safety screening, review the "Will install" list before confirming, and treat unknown marketplaces the way you would treat installing unknown software.

