10 Best Claude Code Plugins for Developers (2026)

Claude Code plugins extend a coding session with reusable skills, language-server diagnostics, and connections to developer services.

They can guide a feature through planning and tests, retrieve documentation for a specific framework version, inspect a running website, or review a GitHub pull request.

In this post, we’ve selected 10 of the best Claude Code plugins for developers working across implementation, UI design, debugging, code review, security, and project memory.

Every entry explains the task it handles and provides an installation command with the relevant setup requirements.

Quick Overview

Plugin	Main Use
Superpowers	Planning, TDD, and debugging
Frontend Design	UI design and implementation
Context7	Current library documentation
TypeScript LSP	Code navigation and diagnostics
Playwright	Browser interaction and testing
Chrome DevTools	Browser debugging and profiling
GitHub	Issues, PRs, and repository operations
Code Review	Automated GitHub PR review
Security Guidance	Security warnings and reviews
Claude Mem	Project memory across sessions

1. Superpowers

Superpowers, created by Jesse Vincent and maintained by Prime Radiant and community contributors, gives Claude a structured software development process. The included skills lead Claude through clarifying the request, approving a design, planning the changes, implementing with tests, and reviewing the result.

After you approve a design, Claude breaks it into concrete tasks with file paths and verification steps. Subagents can implement and review individual tasks, or Claude can work through the plan in the current session. The testing workflow follows a red-green-refactor cycle, and systematic debugging asks Claude to investigate a failure’s cause before proposing a fix.

For a change that touches authentication, data storage, and UI behavior, that process gives the agent an explicit plan to follow and checkpoints before completion. The skills activate when relevant and can guide Claude even when you have not typed a dedicated command.

License: MIT. No required paid Superpowers service. Planning, testing, and review use Claude’s allowance or API budget. Multi-agent execution can consume substantial additional usage. The structured process is most valuable for meaningful changes that justify those extra steps.

/plugin install superpowers@claude-plugins-official

2. Frontend Design

Frontend Design is Anthropic’s plugin for making deliberate visual choices during UI development. It helps Claude establish a direction based on the product, audience, and purpose before implementing the page or application.

Typography, color, layout, and motion follow a design direction tied to the product and its audience. Claude uses that direction to build and refine the UI. A dashboard or landing page can have the correct components yet need more deliberate typography, hierarchy, and details.

Give Claude actual content, audience details, and any existing brand requirements. The skill uses that subject matter to guide the design. It does not add a browser or a component library, so browser inspection requires another tool, such as Playwright or Chrome DevTools.

License: Apache-2.0. No external account, API key, or additional paid service is required.

/plugin install frontend-design@claude-plugins-official

3. Context7

Context7, from Upstash, retrieves current library documentation and code examples for Claude. It earns a place in a development setup when framework changes, renamed APIs, or version differences make remembered examples unreliable.

The plugin connects to Context7’s hosted MCP server. One tool resolves a library name to its documentation identifier, and another retrieves material for the question at hand. Mention the library version in your request when it matters. For example, ask Claude to implement a feature against your installed Next.js version and add use context7.

Access and license: Context7 allows anonymous documentation requests under shared rate limits. Authentication or an API key enables account-based access, and Claude Code can prompt you to sign in. The official marketplace integration is Apache-2.0-licensed. The upstream MCP server code uses MIT, although the hosted documentation backend is proprietary.

/plugin install context7@claude-plugins-official

4. TypeScript LSP

Anthropic’s TypeScript LSP connects Claude Code to a TypeScript and JavaScript language server. Claude gains semantic navigation and diagnostics for a codebase, including definitions, references, and errors.

When you change a shared type or function, finding its references helps establish which callers are affected. Diagnostics give Claude feedback about problems introduced by an edit. This is valuable in larger TypeScript projects, where matching a symbol’s text alone may not identify the intended definition or every relevant use.

Setup: Install both the language server and TypeScript from a terminal, then install the Claude Code plugin. The typescript-language-server executable must be on your PATH. Language analysis does not replace the project’s build or runtime tests.

npm install -g typescript-language-server typescript

/plugin install typescript-lsp@claude-plugins-official

License: Apache-2.0 for the plugin. It runs a local language server and requires no paid documentation or code-analysis service.

5. Playwright

The Playwright plugin connects Claude Code to Microsoft’s Playwright MCP server. It gives Claude a browser for navigating pages, clicking controls, filling forms, handling dialogs, and checking application behavior.

Playwright uses structured accessibility snapshots to identify page elements. Screenshots are also available when appearance matters. A practical request might ask Claude to open a local app, complete a registration form, check its validation messages, and confirm that successful submission reaches the expected page.

Beyond interaction, the server can inspect console messages and network requests, manage tabs, and execute Playwright code. These capabilities help Claude investigate a broken flow and gather evidence about the result. A browser session is exploratory verification. Ask for reusable tests explicitly when you need checks that will run in CI.

Requirements: Use Node.js 20 or newer with npm/npx and a browser that Playwright can launch.

License: Apache-2.0 for the integration and upstream server. No paid browser automation service is required for local use.

/plugin install playwright@claude-plugins-official

6. Chrome DevTools

Google’s Chrome DevTools plugin gives Claude access to a live Chrome browser for debugging and performance analysis. Use it to investigate pages that load slowly, send unexpected network requests, or fail at runtime.

Claude can record a performance trace, inspect network traffic, read console errors, and examine source-mapped stack traces. For example, a slow interaction can be investigated alongside the requests and main-thread activity associated with it. Screenshots and browser automation help reproduce the behavior before inspecting its cause.

There is overlap with Playwright’s browser tools. Playwright is particularly useful for exercising a sequence of interactions and assertions. Chrome DevTools adds a focused route into traces and browser internals when the main question is why the page behaves that way.

Requirements and license: Node.js LTS, npm, and Google Chrome. The plugin and server are Apache-2.0-licensed. Chrome DevTools can inspect and modify content in the browser instance it controls, so avoid exposing unrelated signed-in sessions.

Usage statistics are enabled by default. Set CHROME_DEVTOOLS_MCP_NO_USAGE_STATISTICS=1 before launching Claude Code to opt out. Performance tools may also send trace URLs to Google’s CrUX API. The server’s --no-performance-crux option disables that lookup.

/plugin install chrome-devtools-mcp@claude-plugins-official

7. GitHub

The GitHub plugin connects Claude Code to GitHub’s hosted MCP server. It brings repository context and platform actions into a coding session, including issues, pull requests, code search, and workflow information.

Ask Claude to inspect an issue before implementing it, gather context from a pull request, or investigate a failed GitHub Actions run. With suitable permissions, it can also create or update issues and PRs. This plugin is useful when the context for a change lives in GitHub as well as in the local checkout.

Authentication: The catalog package reads a personal access token from GITHUB_PERSONAL_ACCESS_TOKEN. Set that environment variable before starting Claude Code, and grant the token only the repository access needed for your tasks. Installing the package alone does not authorize access to private repositories.

Access and license: The catalog wrapper is Apache-2.0, and the GitHub MCP server is MIT-licensed. The server is available to GitHub users on all plan types. Tools tied to paid features, such as the Copilot cloud agent, retain those subscription requirements. This remote integration does not require Docker.

/plugin install github@claude-plugins-official

8. Code Review

Anthropic’s Code Review plugin coordinates a review of an existing GitHub pull request. It examines the change through multiple reviewers, then filters potential findings by confidence before posting feedback.

The review considers applicable CLAUDE.md instructions, obvious bugs, git history, comments on previous pull requests, and guidance in code comments. Findings receive confidence scores, and the default workflow filters out scores below 80. That filtering aims to reduce speculative comments, although a high score does not establish that a bug is reproducible.

The command skips closed or draft PRs, changes that do not need review, and PRs it has already reviewed. It is designed for an open pull-request workflow. Claude Code also has a built-in /code-review skill. The installed plugin has its own namespaced /code-review:code-review command for the multi-agent PR review described here.

Requirements and license: An authenticated GitHub CLI (gh auth login), a GitHub repository, and an eligible PR. The source is Apache-2.0-licensed. Multiple review agents consume Claude usage.

/plugin install code-review@claude-plugins-official

Run the command from the repository with the open pull request you want reviewed:

/code-review:code-review

9. Security Guidance

Security Guidance adds security feedback to Claude’s editing and commit workflow. The current Anthropic plugin has three layers: pattern warnings around edits, model-based review of diffs, and an agentic reviewer that investigates commits across related files.

Pattern checks flag risky code, including unsafe deserialization, raw HTML insertion, and hardcoded secrets.
Diff reviews examine the changed code when Claude finishes a turn and return findings to the session.
Commit reviews read related files to investigate issues that depend on data flow across the codebase.

That combination gives it a distinct role alongside Code Review. Security Guidance operates during development and focuses on vulnerability classes such as injection, XSS, SSRF, and authorization failures. The PR review plugin examines a submitted change against repository guidance and historical context.

Requirements: Claude Code 2.1.144 or newer, Python 3.8 or newer, a shell environment, and working model access. The plugin uses the configured Claude API, subscription, or provider connection for its model-based reviews. Those reviews consume model usage in addition to the pattern checks.

License and data: Apache-2.0. Reviews send changed code and, for agentic review, relevant retrieved files to the configured model endpoint. The plugin can miss vulnerabilities or produce false positives, so it supplements human review and security testing.

/plugin install security-guidance@claude-plugins-official

10. Claude Mem

Claude Mem, maintained by Alex Newman, captures development observations and session summaries for later retrieval. It helps carry project decisions and debugging history into a new session when the original conversation is no longer available.

Lifecycle hooks capture activity, a local worker manages processing and search, and a database stores the resulting memories. Claude can search an index, inspect the surrounding timeline, and retrieve the full observations it needs. Relevant context can also be injected when a new session begins.

For example, a later session can retrieve why a previous migration avoided a particular database operation or how an intermittent failure was diagnosed. The web viewer lets you inspect the stored observations. Context settings control what is injected, and <private> tags can exclude sensitive content from storage.

Requirements: Claude Code 2.1.139 or newer and Node.js 20 or newer. The local worker uses Bun and uv for its processing and search dependencies.

License and costs: The open-source plugin is Apache-2.0-licensed. Memory processing can use your Anthropic plan or another configured model provider. The local database stores observations on your computer, but processing through a hosted model or observer sends content to that service.

Add the maintainer’s marketplace and install the plugin. Restart Claude Code after installation:

/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

How to Install Claude Code Plugins

Run the /plugin commands inside an interactive Claude Code terminal session. The first nine entries use Anthropic’s claude-plugins-official marketplace, which Claude Code normally registers at first launch. The Claude Mem marketplace is maintained by its author. If the official marketplace is missing, register it first:

/plugin marketplace add anthropics/claude-plugins-official

Choose an installation scope: user enables the plugin across your projects on that computer, project records it in shared repository settings, and local enables it only for you in that repository. Project settings do not download the plugin onto a teammate’s machine. Each collaborator installs it too.

After installation, check the confirmation message. Some plugins become active immediately. Others require /reload-plugins, a new session, or additional server authentication. Open the Installed tab in /plugin to manage plugins. You can also inspect what is active from the terminal with claude plugin list.

From your system terminal, use claude plugin commands. These examples disable or uninstall Frontend Design:

claude plugin disable frontend-design@claude-plugins-official
claude plugin uninstall frontend-design@claude-plugins-official

Check the commands, skills, hooks, and MCP or LSP servers listed in each plugin’s details before installation. Hooks and local servers can run code under your user account.

For the differences among extension types, see Claude Code Plugins vs Skills vs MCP Servers: What Is the Difference?

Frequently Asked Questions

Are Claude Code plugins free to use?

The 10 plugins here are free to install. Claude Code still requires its own subscription or API access, and plugin activity can consume that allowance. Third-party services, hosted memory, and paid GitHub features can add costs. Open-source plugin code does not establish that a connected service is free or open source.

Do plugins use more context and tokens?

They can. Names and descriptions of model-invocable skills, agents, and commands occupy context while enabled. Their full instructions load when invoked. MCP tool definitions follow Claude Code’s tool-search behavior, and retrieved documents, browser output, and additional review calls contribute usage when performed. A hook itself does not necessarily add prompt tokens, but a hook that calls a model can incur usage.

From your terminal, run claude plugin details with a plugin name to inspect its component inventory and estimated instruction footprint:

claude plugin details superpowers@claude-plugins-official

The estimate does not include runtime model calls or content retrieved by MCP servers. Use /context to inspect the active session.

Can I use these plugins in Claude Code cloud sessions?

Plugins installed on your computer work in the Claude Code environments that use that machine’s plugin settings, including local terminal sessions and supported IDE sessions. Claude Code cloud sessions do not automatically load those local plugins or a repository’s project-level plugin settings. Use a local Claude Code session when you need these installed plugins.