Anthropic launched the official Claude Code plugin marketplace in late 2025, and the ecosystem has grown fast since. What started as a handful of community experiments is now a directory of 33+ official plugins plus hundreds of community ones covering feature development, design, code review, planning, security, and agentic loops.
I use Claude Code daily for vibe coding and longer architecture work, and plugins are what take it from a smart assistant to a structured development workflow. The right ones save real hours per week. The wrong ones eat your context window and add noise.
This guide covers 12 Claude Code plugins worth knowing in 2026, drawn from the official Anthropic marketplace and the community plugins that have earned traction. Each entry covers what the plugin does, how to install it, and what it costs.
12 best Claude Code plugins in 2026
| Plugin | Category | Best For | Pricing |
| feature-dev | Official Anthropic | Guided 7-phase feature development | Free |
| frontend-design | Official Anthropic | Production-grade UI generation | Free |
| /code-review | Official Anthropic | Multi-agent code review with confidence scoring | Free |
| Superpowers | Lifecycle framework | TDD, debugging, brainstorming workflows | Free |
| Deep Trilogy | Planning and implementation | Vague idea to test-driven development | Free |
| Shipyard | Enterprise lifecycle | IaC validation and security auditing | Free |
| Plannotator | Planning visualization | Structured plan annotation and review | Free |
| Local-Review | Local code review | Parallel multi-agent local diff reviews | Free |
| TypeScript/Rust LSP | Language servers | Type checks inside Claude workflows | Free |
| Claude-Mem | Memory | Long-term context across sessions | Free |
| Awesome Claude Plugins | Orchestration | Workflow orchestrator across 500+ services | Free, API key required |
| Ralph Loop | Agentic loop | Autonomous task handling | Free |
1. feature-dev
feature-dev is an official Anthropic plugin with over 89,000 verified installs, which makes it the most popular plugin in the marketplace. It runs a structured seven-phase workflow: discovery, codebase exploration, clarifying questions, architecture design, implementation, quality review, and summary.
What sets it apart is the agent setup. The plugin deploys three specialized sub-agents: code-explorer traces execution paths, code-architect proposes implementation approaches with trade-offs, and code-reviewer catches bugs and convention violations with confidence-scored findings.
Use it for any feature larger than a quick fix. For typo fixes and single-line edits, the plugin skips the full workflow automatically. Anything else, this is the standard starting point.
Key features:
- Seven-phase guided feature development
- Three specialized agents (explorer, architect, reviewer)
- Confidence-scored review findings
- Automatic skip for trivial changes
Setup: /plugin install feature-dev@claude-plugins-official
Pricing: Free.
2. frontend-design
frontend-design is an official Anthropic plugin that addresses a specific failure mode of AI-generated UI. Output from generic prompts tends to converge on similar visual patterns, which makes the result feel templated rather than intentional.
The plugin guides Claude through four design dimensions before any CSS gets written: purpose, tone, constraints, and differentiation. It encourages unusual typography pairings, asymmetric layouts, and varied visual depth, which produces interfaces that read as branded rather than generic.
It pairs naturally with feature-dev on full-stack work. feature-dev runs the architecture and backend, frontend-design handles the visual layer. The combination cuts the back-and-forth iteration cycle that typically eats hours.
Key features:
- Four-dimension design framework
- Unusual typography and layout suggestions
- Mockups with design alternatives
- Pairs naturally with feature-dev
Setup: /plugin install frontend-design@claude-plugins-official
Pricing: Free.
Interesting read: Top Claude Code Alternatives in 2026
3. /code-review
/code-review is an official Anthropic plugin that runs five parallel Sonnet agents on your changes. The agents handle CLAUDE.md compliance (redundant for thoroughness), bug detection in the diff, historical context from git blame, PR history analysis, and code comment review.
The confidence scoring system is the differentiator. Every finding gets a score from 0 to 100, and only issues above the threshold (default 80) make it into the report. This filters the noise that plagues traditional linters and surfaces issues that genuinely matter.
Teams using Claude Code as a final review gate before merge tend to install this one. The combination of multiple reviewers and scored confidence makes the output more useful than a single-pass review.
Key features:
- Five parallel review agents
- Confidence scoring from 0 to 100
- Configurable threshold for noise filtering
- Pre-merge code quality gate
Setup: /plugin install code-review@claude-plugins-official
Pricing: Free.
4. Superpowers
Superpowers combines lifecycle management with a skills framework. Through it, Claude learns brainstorming, subagent development with code review, debugging discipline, TDD, and skill authoring.
The plugin enforces a phase-based planning approach and includes Test-Driven Development as a default rather than an afterthought. The skills framework is composable, so teams can add their own patterns on top of the base workflow.
Teams that already practice spec-driven development find Superpowers maps similar structure into Claude Code without requiring engineers to context-switch between tools.
Key features:
- Structured lifecycle and phase-based planning
- Composable skills with TDD discipline
- Infrastructure as Code validation
- Brainstorming and debugging workflows
Setup: /plugin install superpowers@claude-plugins-official
Pricing: Free.
5. Deep Trilogy
Deep Trilogy is a set of three commands (/deep-plan, /deep-implement, /deep-debug) that automate the path from a rough project idea to test-driven, production-quality code.
The /deep-plan command takes a fuzzy description and produces a structured implementation plan with phases, file changes, and a test strategy. /deep-implement then executes the plan with TDD discipline. /deep-debug handles the inevitable cleanup when something breaks during integration.
Solo developers and small teams without a dedicated planning workflow get the most from this trilogy. It removes the cognitive load of breaking a fuzzy idea into concrete engineering steps.
Key features:
- /deep-plan for structured planning
- /deep-implement for TDD execution
- /deep-debug for systematic cleanup
- Designed for fuzzy-to-finished workflows
Pricing: Free.
6. Shipyard
Shipyard combines the Superpowers skill framework with infrastructure-as-code validation and security auditing. Think of it as Superpowers for teams shipping to production with compliance requirements.
The plugin supports IaC validation for Terraform, Ansible, Docker, Kubernetes, and CloudFormation. It includes a dedicated auditor agent for security checks and a code simplification pass that catches cross-task duplication and AI-generated bloat.
Enterprise teams and anyone building infrastructure-heavy systems get the most value here. Shipyard adds production-grade safeguards that Superpowers alone does not cover.
Key features:
- IaC validation across major platforms
- Dedicated security auditor agent
- Cross-task duplication detection
- Code simplification pass
Setup: Available at shipyard.build/agents/claude-code.
Pricing: Free.
7. Plannotator
Plannotator improves Claude Code’s planning mode by providing a structured way to manage, annotate, and share plans. It includes a UI for reviewing and approving plans before execution, similar to artifact-style review patterns.
The workflow gap this closes is real. Teams that need plans reviewed by humans before Claude executes get a visible, shareable artifact rather than a buried Claude Code session.
Solo developers running short tasks find this heavier than needed. The value scales with team size and project complexity.
Key features:
- Annotated planning mode
- UI for plan review and approval
- Shareable plans across team members
- Visual structure for complex tasks
Setup: Available at plannotator.ai.
Pricing: Free.
8. Local-Review
Local-Review runs five agents in parallel on uncommitted local changes. Each agent reviews the diff from a different angle, and the plugin only flags issues scored 80 or above to filter noise.
The use case is the second opinion before commit. Developers who want a thorough review before pushing get immediate feedback without waiting for a PR review queue. The plugin works on local git diffs directly.
The honest trade-off is cost. Running five agents in parallel burns through tokens fast, especially on larger diffs. Worth using selectively on critical changes rather than every commit.
Key features:
- Five parallel review agents
- Threshold-based issue filtering
- Works on local git diffs
- Immediate pre-commit feedback
Pricing: Free. Token costs scale with diff size.
9. TypeScript and Rust LSP Plugins
These plugins integrate Language Server Protocol checks directly into Claude Code. For TypeScript and Rust projects, this surfaces real type errors, lint warnings, and compiler diagnostics inside the Claude workflow.
The value lives in the feedback loop. Without LSP integration, Claude generates code that may compile in theory but fails type checks once you run the build. With LSP wired in, Claude sees type errors immediately and corrects course before you spot them.
Teams running TypeScript or Rust at scale should treat this as close to essential. The overhead is minimal and the time saved on type error debugging adds up fast.
Key features:
- LSP integration for TypeScript and Rust
- Real type checks inside Claude Code
- Compiler diagnostics surfaced in workflow
- Tight feedback loop for typed languages
Pricing: Free.
10. Claude-Mem
Claude-Mem adds long-term memory to Claude Code. It captures and compresses Claude’s actions across sessions, then injects relevant context into future sessions automatically.
The plugin uses SQLite and Chroma for hybrid search with vector embeddings. Privacy controls let you exclude specific information from storage, which matters for teams working on sensitive codebases.
Developers working on long-running projects benefit the most. Claude remembers what you discussed last session, which patterns you prefer, and what context matters for the current work, which removes the constant re-explanation.
Key features:
- Long-term memory across sessions
- SQLite plus Chroma hybrid search
- Vector embeddings for relevant context
- Privacy controls for excluded content
Setup: Repo: github.com/thedotmack/claude-mem.
Pricing: Free and open source.
11. Awesome Claude Plugins
Awesome Claude Plugins is a curated registry maintained by ComposioHQ that turns Claude Code into a workflow orchestrator. Through the Composio Tool Router, Claude can interact with hundreds of external services from a single endpoint.
Three skills inside the registry stand out. The connect-apps skill links Claude to 500+ SaaS applications in one command. The ship skill handles full PR automation from linting through deployment. The mcp-builder skill guides you through designing your own agent-optimized MCP servers.
Teams that need Claude to coordinate work across many tools get the most from this registry. The Composio Tool Router minimizes context window usage by loading only the tools Claude actually calls.
Key features:
- Curated registry with custom slash commands
- 500+ SaaS integrations via connect-apps
- ship skill for full PR automation
- Composio Tool Router for context efficiency
Setup:
git clone https://github.com/ComposioHQ/awesome-claude-plugins.git
cd awesome-claude-plugins
claude –plugin-dir ./connect-apps
Free API key from platform.composio.dev.
Pricing: Free. Composio API key required.
12. Ralph Loop
Ralph Loop is an agentic loop plugin for autonomous task handling. Rather than single-shot sessions, Ralph Loop runs continuously, taking action, observing the result, and deciding the next step until the task is complete.
The original Ralph Wiggum variant was built for visual testing of iOS apps via the Xcode simulator. The broader Ralph Loop framework now supports general agentic workflows where Claude needs to operate autonomously for extended periods.
The honest warning is token cost. Composio’s writeup flagged a developer who wiped out a $20 plan in a single Ralph Loop run. Use it for tasks that genuinely need autonomous iteration, not for one-off scripts.
Key features:
- Continuous agentic loop
- Visual testing via Xcode MCP (original)
- General autonomous task handling
- Extended-session workflows
Setup: /plugin install ralph-loop@claude-plugins-official
Pricing: Free. Token costs scale with loop duration.
How to choose the right Claude Code plugins for your workflow
Start with the official Anthropic plugins. feature-dev, frontend-design, and /code-review cover the most common workflows. They are free, well-maintained, and battle-tested by Anthropic. For most developers, these three handle 80% of the value the plugin marketplace offers.
From there, add plugins based on the specific friction in your workflow. Complex planning workflows benefit from Deep Trilogy or Plannotator. Infrastructure-heavy systems benefit from Shipyard. TypeScript or Rust projects benefit from the LSP plugins. Long-running projects benefit from Claude-Mem for persistent context.
The trap is installing every plugin in the marketplace because they sound useful. Each plugin adds slash commands and context overhead. Pick three to five that match your actual work and resist the urge to bolt on more.
Frequently asked questions
What are Claude Code plugins?
Claude Code plugins are extensions that add slash commands, skills, and agents to Claude Code sessions. They run inside Claude Code itself, unlike MCP servers which expose external tools.
How do I install Claude Code plugins?
Most official plugins install via the Anthropic marketplace using /plugin install <name>@claude-plugins-official. Community plugins typically install via git clone and claude –plugin-dir ./<plugin>.
Are Claude Code plugins free?
Most official Anthropic plugins are free. Some community plugins are free but burn tokens fast (Local-Review, Ralph Loop). Token costs scale with plugin usage.
What is the difference between plugins, skills, and MCP servers?
Plugins are slash command extensions inside Claude Code. Skills are reusable patterns Claude follows during execution. MCP servers expose external tools to Claude. For a deeper look at the MCP layer, see our guide to the best MCP servers for Claude Code.
Which Claude Code plugins are official?
Anthropic maintains an official plugin marketplace at the claude-plugins-official repo. Official plugins include feature-dev, frontend-design, /code-review, and others.
