
Anthropic official plugins refer to Anthropic-developed plugins — distribution packages that extend Claude Code with Skills, Hooks, MCP servers, and more — that are developed and maintained by Anthropic itself.
This article is intended for both developers already using Claude Code and those considering team adoption. It provides an overview of the official plugins and clarifies what claude-code-setup actually is as an entry point into that ecosystem. By the end, you will have a bird's-eye view of the 35 Anthropic-developed plugins available at the time of writing, along with a framework for distinguishing between "official" and "Anthropic-developed" without conflating the two.
Plugins are extension packages for Claude Code. First, it is worth understanding what can be packaged into a plugin and where plugins are distributed from.
A single plugin is a "box" that bundles multiple components together. The official documentation defines the components that can be included in a plugin as follows:
| Directory/File | Role |
|---|---|
skills/ | Specialized procedures and knowledge that Claude invokes based on context |
commands/ | Slash commands (flat Markdown; new plugins are recommended to use skills/ instead) |
agents/ | Definitions for sub-agents (agents for specialized tasks) |
hooks/ | Automated processing triggered by events such as PreToolUse / PostToolUse |
.mcp.json | Configuration for MCP servers (external tool integrations) |
.lsp.json | Configuration for LSP servers (code intelligence) |
monitors/ | Configuration for background monitoring of logs and files |
bin/ | Executables added to the Bash PATH when the plugin is enabled |
settings.json | Default settings applied when the plugin is enabled |
Not all components need to be included. You can create a small plugin with just a single slash command, or a larger plugin that bundles multiple components together.
Plugins are distributed through marketplaces. Anthropic operates two public marketplaces.
claude-plugins-official is the official marketplace curated by Anthropic and is automatically available when Claude Code is installed. The other, claude-community (claude-plugins-community), is a community-oriented marketplace where third-party submissions are listed after review; it can be added with /plugin marketplace add anthropics/claude-plugins-community and used as @claude-community.
Furthermore, the official repository itself is divided into two layers. The /plugins directory contains Anthropic-developed plugins, while /external_plugins contains plugins from partners and the community. When this article uses the term "Anthropic-developed," it refers to the former.
It is not practical to try every plugin in the official marketplace one by one. claude-code-setup is an "entry point" plugin that narrows things down to only those that will genuinely be effective for your project.
claude-code-setup is a tool that proposes configurations, not one that rewrites them. The official documentation explicitly states: "This skill is read-only - it analyzes but doesn't modify files."
Specifically, it analyzes package.json, the programming language in use, directory structure, dependencies, and code patterns, then presents automations that would be of high value to introduce into the project. The decision of whether to adopt a suggestion, and the actual configuration work itself, remain entirely in the user's hands. Misunderstanding this point easily leads to the misconception that "nothing changed even after installing the official plugin."
claude-code-setup provides recommendations across the following 5 categories. Since it generally narrows down to the top 1–2 items per category, the suggestion list never grows so large that it becomes difficult to choose from.
| Category | Suggestion Content | Examples |
|---|---|---|
| MCP Servers | Integration with external tools and services | Playwright for React, Context7 for documentation search |
| Skills | Reusable workflows and specialized knowledge | Running tests, creating PRs, frontend design assistance |
| Hooks | Automated processing triggered by events | Formatting after save, warnings for dangerous file changes, running lint |
| Subagents | Sub-agents for specialized review | Security, performance, accessibility, and quality reviews |
| Slash Commands | Turning common tasks into commands | /test, /pr-review, /explain, etc. |
When you want to explore a specific category in depth, you can request something like "dig deeper into MCP only," and you will receive a focused deep-dive into that category.
Usage is as simple as making a request in conversation. On Claude Code, ask something like "recommend automations for this project," "help me set up Claude Code," or "what hooks should I use?" and claude-code-setup will read your project and return its recommendations.
There are two ways to install it. You can run /plugin in Claude Code and search from the Discover tab, or you can specify it directly via command.
/plugin install claude-code-setup@claude-plugins-official
The format {plugin-name}@{marketplace-name} is common to all official plugins, and claude-plugins-official is available from the start, as mentioned earlier.
What claude-code-setup typically proposes are the Anthropic-made plugins introduced below. At the time of writing, the /plugins directory of the official repository contains 35 Anthropic-made plugins. Familiarizing yourself with them by role will help you make faster decisions when you receive a suggestion.
The plugins you will reach for most often in day-to-day development are those for workflows and code quality.
| Plugin | Role |
|---|---|
| claude-code-setup | Analyzes the project and proposes automation configurations |
| claude-md-management | Audits CLAUDE.md quality and updates project memory |
| session-report | Converts session token and cache efficiency data into an HTML report |
| feature-dev | Feature development workflow including exploration, design, and quality review |
| code-review | Automatically reviews PRs using multiple specialized agents |
| pr-review-toolkit | PR review focused on comments, tests, type design, and more |
| code-simplifier | Reorganizes code for readability while preserving functionality |
| commit-commands | Git workflows such as commit, push, and PR creation |
| code-modernization | Modernization support for legacy code such as COBOL and older Java/C++ |
| security-guidance | Warns about dangerous code patterns during editing |
| frontend-design | Assists with frontend and UI implementation with high design quality |
The rest are for learning purposes, specialized tasks, plugin developers, and per-language LSPs.
| Plugin | Role |
|---|---|
| playground | Generates an interactive single-file HTML playground |
| explanatory-output-style | Adds educational explanations of implementation decisions and code structure |
| learning-output-style | A learning mode that prompts code contributions at key decision points |
| math-olympiad | Solves competitive mathematics problems with a verification agent |
| agent-sdk-dev | Supports development of Claude Agent SDK applications |
| plugin-dev | Supports Skills, Agent, and validation for plugin creation |
| skill-creator | Creates, improves, and evaluates new Skills |
| mcp-server-dev | Supports design, implementation, and deployment of MCP servers |
| hookify | Creates custom Hooks from conversation patterns and rules |
| ralph-loop | A self-referential loop that iteratively improves the same task |
| cwc-makers | Sets up Code with Claude Makers Cardputer |
| example-plugin | A reference implementation of plugin structure |
LSP (Language Server Protocol) plugins give Claude code intelligence capabilities such as jump-to-definition, reference search, and type error detection. Anthropic's own plugins alone cover 12 languages: TypeScript/JavaScript (typescript-lsp), Python (pyright-lsp), Rust (rust-analyzer-lsp), C/C++ (clangd-lsp), C# (csharp-lsp), Go (gopls-lsp), Java (jdtls-lsp), Kotlin (kotlin-lsp), Lua (lua-lsp), PHP (php-lsp), Ruby (ruby-lsp), and Swift (swift-lsp). The LSP plugin for your project's primary language is a group worth installing from the very start.
Official plugins are convenient, but without a precise understanding of what "official" actually means, you risk making poor trust judgments.
Some entries in Claude's plugin directory are labeled "Anthropic Verified" or "Made by Anthropic." However, being listed in the official marketplace claude-plugins-official does not in itself mean "made by Anthropic."
claude-plugins-official is curated by Anthropic, but its scope includes plugins from partners and the community (found under /external_plugins in the official repository). Inclusion is at Anthropic's discretion, and there is no application process. Conversely, submitting through the community marketplace submission form does not result in a plugin being added to the official marketplace.
If you want to verify strictly whether something was made by Anthropic, the reliable method is to check whether it exists in the /plugins directory of the official repository anthropics/claude-plugins-official. The 35 plugins counted in this article are based on that criterion.
Another pitfall is treating claude-code-setup as a "button that finishes configuration." As noted above, it is read-only, and what it produces are merely suggestions. Deciding which ones to adopt, and whether they fit your organization's security policies and review processes, remains a matter of human judgment.
The same applies to the safety of plugins themselves. The official documentation cautions that "Anthropic does not manage the MCP servers, files, or software contained in plugins." Install only after confirming that the provider is trustworthy. This is a universal principle that applies equally to Anthropic-made and third-party plugins alike.
What you should do first depends on your role.
If you are a developer, the quickest starting point is to ask claude-code-setup to "recommend automations for this project" on your local repository. Since it is read-only, you can get a sense of "what would work for this project" without breaking anything. From the suggestions, install the LSP plugin for your primary language along with one or two everyday tools like commit-commands or code-review, confirm their effectiveness, and then expand from there. There is no need to install everything at once.
If you are evaluating a team-wide rollout, what you should focus on is "what to standardize." Using claude-code-setup's suggestions as a starting point, prioritize in order of team impact: use code-review and pr-review-toolkit to raise review quality, use security-guidance to catch dangerous changes, and use claude-md-management to prevent project knowledge from degrading. The marketplace can be configured per repository and can also be hosted in a private repository for internal use, allowing you to distribute a decided configuration as a team standard. The guiding question should be not "is it official?" but rather "where is our development workflow bottlenecked?"
Finally, let me address three common questions.
No, it won't be overwritten. claude-code-setup is read-only — it simply analyzes your package.json and directory structure to generate suggestions. Any file changes or actual installation of suggested plugins must be performed explicitly by the user. "Analyze but don't modify" is a core design principle of this plugin.
The most reliable way to distinguish plugins is by their source directory, not their display label. Anthropic-made plugins are located in /plugins of the official repository anthropics/claude-plugins-official, while partner- and community-sourced plugins are in /external_plugins of the same repository. Being listed as "Anthropic Verified" or appearing in the official marketplace does not prove that a plugin was made by Anthropic.
Marketplaces can be configured at the repository level, and it is also possible to host a marketplace in a private repository for internal use. By using claude-code-setup's suggestions to establish team standards and distributing common rule-based plugins such as code-review and security-guidance as a foundation, it becomes easier to keep Claude Code's behavior consistent across team members.

Anthropic official plugins are a suite of Anthropic-developed packages that extend Claude Code via Skills, Hooks, MCP, LSP, Monitors, and more — with 35 plugins available in /plugins of the official repository at the time of writing. The entry point to these is claude-code-setup, which analyzes your project in read-only mode and suggests only the plugins likely to be effective, organized across five categories: MCP, Skills, Hooks, Subagents, and Slash Commands.
Two points are critical. First, this is a suggestion tool, not an auto-configuration tool. Second, being listed on the official marketplace or displaying a "Verified" badge is not synonymous with being made by Anthropic. Keep these two points in mind, and you'll never be at a loss — whether you're receiving suggestions or making trust decisions.
Even with a clear picture of the full official plugin ecosystem, designing where in your development workflow to apply which automations is a separate undertaking. Unimon supports organizations in planning and implementing AI tools, including Claude. If you're considering adopting Claude Code as a team or building a development workflow that incorporates AI, feel free to reach out.

Yusuke Ishihara
Started programming at age 13 with MSX. After graduating from Musashi University, worked on large-scale system development including airline core systems and Japan's first Windows server hosting/VPS infrastructure. Co-founded Site Engine Inc. in 2008. Founded Unimon Inc. in 2010 and Enison Inc. in 2025, leading development of business systems, NLP, and platform solutions. Currently focuses on product development and AI/DX initiatives leveraging generative AI and large language models (LLMs).