The Claude Code Plugins Worth Installing in 2026

Commissioned, Curated and Published by Russ. Researched and written with AI.

What’s New This Week

No material updates – this is the initial publication.

Changelog

Claude Code out of the box is a capable coding agent. With the right plugins it becomes a different tool. The plugin system – launched by Anthropic in late 2025 – lets you install collections of slash commands, specialized agents, hooks, and MCP servers with a single command. They show up in your session and run automatically or on demand.

Install any plugin with:

/plugin marketplace add {publisher}/{plugin-name}
/plugin install {plugin-name}

The official Anthropic marketplace (claude-plugins-official) is the most reliable source. Third-party plugins are available from community publishers and installable the same way.

Here is what is actually worth installing.

Must-Install

context7

Publisher: Upstash (official marketplace)

The most important plugin in this list. Claude Code’s training data has a cutoff. It does not know your current library versions. Context7 solves this: it is an MCP server that fetches up-to-date, version-specific documentation and injects it directly into your prompts. When you reference a library, Context7 resolves it to a specific version and pulls current API docs and code examples.

The practical effect is fewer hallucinated APIs and deprecated patterns. If you have ever watched Claude confidently write code against a method that was removed six months ago, this is the fix.

Context7 integrates via a SessionStart hook so it is available throughout your session without any manual prompting.

/plugin marketplace add context7
/plugin install context7

Workflow-Specific

These plugins are high-value for specific types of work. If your work matches the use case, install them. If it does not, skip.

deep-plan

Publisher: piercelamb-plugins

/deep-plan takes a requirements file – as vague or as detailed as you like – and produces a structured implementation plan. It runs codebase research, conducts an interview to clarify requirements, and performs a review phase that can use an external LLM (Gemini or OpenAI) if you provide an API key, or falls back to a subagent. The output is a set of section files organized by component.

Designed for engineers who want to understand what they are building before Claude starts writing it. The plugin’s own documentation describes it as targeted at users who “prefer deep planning before implementation” and are not operating under strict token constraints. That is an honest description. This is not lightweight.

Slash command: /deep-plan @planning/spec.md

deep-project

Publisher: piercelamb-plugins

The entry point to the pierce-lamb trilogy. Takes a broad, vague idea and decomposes it into components that are each ready to be handed to /deep-plan. If you know what you want to build but have not broken it down yet, start here.

Slash command: /deep-project @planning/requirements.md

deep-implement

Publisher: piercelamb-plugins

The final stage. Takes the section files produced by /deep-plan and implements them with TDD, code review, and git workflow. Picks up from where planning left off and works through sections in order. State is managed through the filesystem so interrupted sessions can resume.

The three piercelamb plugins form a pipeline: decompose with /deep-project, plan with /deep-plan, build with /deep-implement. Each is independent – you can use any of them without the others – but they are designed to work in sequence.

Slash command: /deep-implement @planning/sections/

code-review

Publisher: Anthropic (official)

Automated PR review using five parallel Sonnet agents. Each agent specialises in a different dimension: CLAUDE.md compliance, bug detection, historical context, PR history, and code comments. Results are scored by confidence to filter noise before they reach you.

Slash command: /code-review

Worth installing for teams where PRs move fast and you want a consistent first pass before human review.

security-guidance

Publisher: Anthropic (official)

Hook-based, not command-based. Installs a SessionStart hook that injects security reminders at the beginning of each session. No slash command to invoke – it runs automatically and keeps security considerations in context while Claude is writing or modifying code.

This is a light-touch plugin. It does not run an audit or produce a report. It functions more as a persistent nudge that keeps Claude oriented toward secure patterns throughout the session.

Memory and State

claude-mem

Publisher: thedotmack (community)

Claude Code sessions are stateless by default. Start a new session and the previous one is gone. claude-mem fixes this: it automatically captures everything that happens during your coding sessions, compresses it with AI, and injects relevant context back into future sessions. Per-project CLAUDE.md files are generated automatically to record activity timelines for each folder.

The install process sets up a background worker service and registers the plugin hooks. The plugin’s documentation notes that installing via npm (npm install -g claude-mem) does not register the hooks – always install through the plugin system.

Slash commands: /claude-mem-status, /claude-mem-feed

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

claude-md-management

Publisher: Anthropic (official)

Two complementary tools for keeping your CLAUDE.md and AGENTS.md files useful:

• claude-md-improver – a skill that audits CLAUDE.md against the current codebase state. Invoke it when you want to check alignment: “audit my CLAUDE.md files.”
• /revise-claude-md – a command that captures learnings from the current session and writes them back into your CLAUDE.md. Run it at the end of sessions where Claude discovered something worth preserving.

Between these two, your CLAUDE.md stays current rather than becoming a document that drifts away from the actual codebase.

Nice to Have

superpowers

Publisher: Anthropic (official, based on obra/superpowers)

A structured skills framework that teaches Claude better engineering habits: brainstorming before coding, TDD, subagent-based code review, and debugging discipline. Skills trigger automatically at relevant points in the session without you needing to invoke them manually. The practical effect is an agent that pauses to plan and test rather than charging straight into implementation.

Token usage is higher than a vanilla session. That is the honest tradeoff.

commit-commands

Publisher: Anthropic (official)

Streamlined git operations. Three commands:

• /commit – stage and commit with a generated message
• /commit-push-pr – commit, push, and open a PR in one step
• /clean_gone – clean up local branches that no longer have a remote tracking branch

Simple utility. If you run these operations frequently, the shortcuts add up.

code-simplifier

Publisher: Anthropic (official)

An agent-based plugin for simplifying existing code. The use case is refactoring: point it at code that has grown complex and let it suggest a cleaner version. Specific slash commands are not documented in the plugin’s public repository – the simplification logic runs through agents.

frontend-design

Publisher: Anthropic (official)

A skill that auto-invokes whenever Claude is working on frontend code. It provides guidance toward distinctive, production-grade interfaces rather than generic AI-generated aesthetics – opinionated direction on typography, animation, and visual detail. If your frontend output tends toward bland, this nudges Claude toward more considered design choices.

Installing and Managing Plugins

Browse and search the marketplace:

/plugin

Install from the official Anthropic marketplace:

/plugin install {plugin-name}@claude-plugins-official

Install from a third-party publisher:

/plugin marketplace add {publisher}/{plugin-name}
/plugin install {plugin-name}

List installed plugins:

/plugin list

Enable or disable without uninstalling:

/plugin enable {plugin-name}
/plugin disable {plugin-name}

Finding More

The official marketplace is at claude.com/plugins and the source repository is at github.com/anthropics/claude-plugins-official.

Community-curated lists worth bookmarking:

• github.com/hesreallyhim/awesome-claude-code
• github.com/ccplugins/awesome-claude-code-plugins

The plugin ecosystem is moving fast. New plugins appear regularly. The official marketplace has a quality bar; third-party sources do not. Read the README and check repository activity before installing anything from an unknown publisher.

The four plugins I would put in every setup without qualification: context7, claude-mem, claude-md-management, and superpowers. Everything else depends on what you are building.