Claude Code Powerkit
Neens6655/claude-code-powerkit
Setup completo do Claude Code em uma pasta — sem terminal necessário
Ver no GitHubUm MCP server que automatiza a configuração completa do Claude Code em menos de 5 minutos, instalando 10 servidores MCP, 54 skills e 70 subagents especializados. Tudo funciona sem terminal — basta colar uma URL no Claude Code e deixar que ele configure automaticamente.
Instalação
npx convez-ai Neens6655/claude-code-powerkit1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Acelerar onboarding de desenvolvimento com Claude Code
- Configurar ambiente de trabalho com E2E testing, browser automation e 70 agents especializados
- Automatizar instalação de dependências, skills e MCP servers
Veredito

Excelente para quem quer sair do zero com Claude Code em minutos, sem lidар com linhas de comando. Abordagem criativa (Claude instalando Claude), mas depende fortemente da execução confiável do bootstrap automático.
README
A tradução para o português deste README ainda está na fila.
Claude Code Powerkit
One paste. Full setup. Skip the first week.
The fastest way to go from a blank Claude Code install to a fully equipped development environment — 10 MCP tools, 54 skills, and 70 specialist subagents configured in under 5 minutes, no terminal required.
The mechanic: Paste one URL into Claude Code. Claude reads the installer, fetches every skill and agent from GitHub, writes your config files, and prints a verified summary. Claude installs Claude.
Quick Start
Step 1 — Install prerequisites (Node.js is the only hard requirement):
# Install Homebrew if you don't have it
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
# Required
brew install node git
# Recommended (faster installs, GitHub automation)
brew install bun gh
# Optional (needed for specific advanced skills)
brew install python@3.12 ffmpeg
# Required
winget install OpenJS.NodeJS.LTS
winget install Git.Git
# Recommended
winget install Oven-sh.Bun
winget install GitHub.cli
# Optional
winget install Python.Python.3.12
winget install Gyan.FFmpeg
After installing, close and reopen your terminal.
# Node.js v20
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs git
# Recommended
curl -fsSL https://bun.sh/install | bash
sudo apt install gh
# Optional
sudo apt-get install -y python3 ffmpeg
Verify you're ready:
node --version # v20.x.x or higher
npx --version # any version number
Step 2 — Open Claude Code and paste this:
Set me up using https://raw.githubusercontent.com/Neens6655/claude-code-powerkit/master/BOOTSTRAP.md
Claude does everything else. Takes 3–5 minutes.
Step 3 — First test (no API key needed):
Take a Playwright screenshot of google.com
If that works, you're set up correctly.
What Gets Installed
| What | Count | |
|---|---|---|
| 🔌 | MCP Servers | 10 |
| ⚡ | Curated Skills | 30 |
| 🌐 | Composio Community Skills | 24 |
| 🤖 | Specialist Subagents | 70 |
| 🎨 | Built-in Design System | 300+ CSS tokens |
| ⚙️ | Settings + Hooks | npm/yarn block, build gate, quality reminder |
🔌 MCP Servers
MCP servers extend what Claude can do beyond the chat window. Two work immediately with no key. The rest unlock as you add API keys.
| Server | What it unlocks | Key required |
|---|---|---|
| Playwright | Take screenshots, control browsers, run E2E tests, scrape any page | ✅ No key — works now |
| Memory | Remember facts, preferences, and context across sessions permanently | ✅ No key — works now |
| Brave Search | Real-time web search — current events, docs, prices, competitors | Brave API — 2,000/mo free |
| GitHub | Create PRs, push files, open issues, search code, manage repos | GitHub token — free |
| Perplexity | AI-powered research — synthesizes multiple sources into clean answers | Perplexity API |
| Firecrawl | Convert any website to clean markdown or structured JSON | Firecrawl — 500 pages/mo free |
| Apify | 4,000+ pre-built scrapers: Instagram, LinkedIn, Amazon, TikTok, and more | Apify token — free tier |
| Google Maps | Places search, geocoding, directions, distance calculations | Maps API — $200 free credit |
| OpenRouter | 100+ AI models from one key: GPT-4o, Gemini 2.5, Llama, Mistral, and more | OpenRouter — free models available |
| Replicate | 50,000+ AI models for image, video, audio, 3D generation | Replicate — credits on signup |
⚡ Skills — Full Catalog
Skills are slash commands that give Claude deep expertise in a specific domain. Type /skill-name to activate.
🎓 Learn While You Build
| Skill | What it does | Example |
|---|---|---|
/sensei | Before every code task, names the CS concept at play, asks you to predict the outcome, then teaches and builds. Kill switch: #noteach | Write a debounce function → Claude names the concept, asks a question, waits for your answer |
🏗 Planning & Architecture
| Skill | What it does | Example |
|---|---|---|
/prd | Interview-driven product requirements doc — Claude asks 8 questions then writes the full spec | /prd → describe your idea when prompted |
/architect | System design, file maps, data models, ADRs, risk assessment | /architect a multi-tenant SaaS with Stripe billing |
/kickoff | Chains architect → design → build → code review with approval gates at each step | /kickoff build a user dashboard with analytics |
/app-builder | Scaffold any app type from a one-sentence description — 13 project templates | /app-builder a REST API with auth and PostgreSQL |
/autopilot | Brief → deployed product, zero manual gates. Claude plans, builds, tests, and ships | /autopilot a landing page for my SaaS |
🧠 AI Advisory Council
| Skill | What it does | Example |
|---|---|---|
/plato | Spawn 5 expert advisors (Strategist, Architect, Customer, Contrarian, QA) to debate any decision. Get a synthesis verdict with action items | /plato quick should I build this as a SaaS or a one-time tool? |
/plato quick | Fast 3-advisor check — findings only, under 2 minutes | /plato quick review my pricing page |
🎨 Design & Visual
| Skill | What it does | Example |
|---|---|---|
/design | Built-in design system — 7 style modes, production React + Tailwind output, WCAG compliant. No setup. | /design a pricing page with 3 tiers |
/design editorial | Magazine-style editorial layout | /design editorial a blog post page |
/design glass | Modern SaaS dark glass aesthetic | /design glass a dashboard sidebar |
/design brutalist | Raw, confrontational anti-design | /design brutalist a portfolio homepage |
/canvas-design | Algorithmic visual art to PNG/PDF | /canvas-design a geometric poster with my brand colors |
/data-storytelling | Transform raw data into compelling narrative + visualization | /data-storytelling my monthly revenue CSV |
/scroll-experience | Cinematic scroll storytelling with Lenis + Framer Motion / GSAP | /scroll-experience a product launch page |
/algorithmic-art | Generative art with p5.js — particles, geometry, noise fields | /algorithmic-art a flowing particle system |
/video-forge | TTS-narrated video pipeline: script → voice clone → Remotion render → localize (advanced — requires ffmpeg + Remotion + voice API) | /video-forge build a 60s product promo |
✅ Code Quality & Debug
| Skill | What it does | Example |
|---|---|---|
/fix | Trace root cause, explain in plain English, fix scoped to the broken area | /fix TypeError: Cannot read property 'map' of undefined |
/explain | Any code, file, or error explained in plain English — layered depth | /explain what does this regex do |
/vibe-code-auditor | Catch fragile AI-generated code before it ships — finds hardcoded values, missing error handling, hidden assumptions | /vibe-code-auditor on any file |
/iterate | Audit-first improvement for shipped products — finds issues before fixing | /iterate my checkout flow is slow |
/feedback-loop | Capture a correction → extract a rule → save it so future sessions don't repeat the mistake | /feedback-loop never use inline styles |
✍️ Writing & Content
| Skill | What it does | Example |
|---|---|---|
/beautiful-prose | Hard-edged writing style contract — kills generic AI voice, enforces specific, concrete language | /beautiful-prose write a product announcement |
/marketing-psychology | Behavioral science copy — mental models, social proof, loss aversion, conversion optimization | /marketing-psychology write a landing page hero |
/baoyu-infographic | Research-backed infographics in 18 visual styles | /baoyu-infographic the history of AI |
/baoyu-image-gen | AI image generation across Replicate, Stable Diffusion, DALL-E | /baoyu-image-gen a Bauhaus-style product shot |
💻 Code Expertise
| Skill | What it does | Example |
|---|---|---|
/react-best-practices | 40+ Vercel-level performance rules — Server Components, bundle splitting, suspense boundaries | Use before building any React component |
/tailwind-patterns | Tailwind v4 CSS-first patterns — @theme {} in CSS, not config files | Use for any Tailwind v4 project |
/typescript-expert | Deep TypeScript — narrowing, generics, strict patterns, discriminated unions | /typescript-expert make this type-safe |
🚀 Advanced & Automation
| Skill | What it does | Example |
|---|---|---|
/loki-mode | Autonomous multi-agent startup system — spins up coordinated agents to build in parallel | /loki-mode build a SaaS MVP in one session |
/dispatching-parallel-agents | Run 2+ independent tasks simultaneously — cuts build time in half | Run in parallel: build the header AND the API |
/mcp-builder | Build a custom MCP server from scratch — write tools Claude can call | /mcp-builder a server that reads my Notion database |
/micro-saas-launcher | End-to-end SaaS launch framework — from idea to live product | /micro-saas-launcher my idea is X |
/cheap-route | Route non-critical tasks (docs, summaries, formatting) to cheaper models via OpenRouter to save quota | /cheap-route explain this function |
🌐 Composio Community Skills (24 additional skills)
Installed automatically from ComposioHQ/awesome-claude-skills:
| Skill | What it does |
|---|---|
/artifacts-builder | Multi-component HTML/React artifact creation |
/brand-guidelines | Apply brand standards to any content |
/changelog-generator | Auto-generate release notes from git commits |
/competitive-ads-extractor | Analyze competitor advertising strategies |
/content-research-writer | Research-backed writing with citations |
/developer-growth-analysis | Dev productivity and growth insights |
/document-skills | Document processing workflows |
/domain-name-brainstormer | Generate and validate domain name ideas |
/file-organizer | Intelligent file organization |
/image-enhancer | Image quality improvement workflows |
/internal-comms | Company communication templates |
/invoice-organizer | Tax-ready invoice management |
/lead-research-assistant | Lead qualification and outreach |
/meeting-insights-analyzer | Meeting pattern and insights analysis |
/raffle-winner-picker | Cryptographically secure winner selection |
/skill-creator | Build new Claude skills from scratch |
/skill-share | Package and share your custom skills |
/slack-gif-creator | Slack-optimized GIF generation |
/tailored-resume-generator | Job-matched resume creation |
/theme-factory | Font and color theme presets |
/twitter-algorithm-optimizer | Tweet optimization for engagement |
/video-downloader | YouTube and platform video downloads |
/webapp-testing | Playwright-based web app testing |
/template-skill | Skill scaffolding template |
🤖 Specialist Subagents (70 total)
Subagents are experts Claude spawns automatically when a task needs deep domain knowledge. You don't invoke them directly — Claude routes to the right one based on what you're building.
How it works: Ask Claude to "build me a FastAPI endpoint with auth" and it automatically delegates to the fastapi-expert and security-auditor agents in parallel. You get specialist-level output without knowing the agent names.
| Agent | What it does |
|---|---|
product-manager | Product strategy, roadmap, prioritization frameworks |
content-marketer | Content strategy, SEO, distribution playbooks |
ux-researcher | User research, interview scripts, usability analysis |
business-analyst | Requirements gathering, process mapping, gap analysis |
project-manager | Agile planning, sprint design, delivery tracking |
wordpress-master | WordPress development, plugins, performance optimization |
research-analyst | Comprehensive research across any domain |
project-idea-validator | Brutal go/no-go product validation — kills bad ideas fast |
competitive-analyst | Competitive intelligence, positioning analysis |
market-researcher | Market sizing, consumer insights, trend analysis |
frontend-developer | React, Vue, Angular UI specialist |
fullstack-developer | End-to-end feature development |
ui-designer | Visual design and interaction patterns |
debugger | Advanced debugging and root cause analysis |
code-reviewer | Code quality, patterns, anti-patterns |
qa-expert | Test strategy, test automation, coverage analysis |
ai-writing-auditor | Detect and rewrite AI-sounding patterns |
technical-writer | Technical documentation, guides, API docs |
| Agent | What it does |
|---|---|
architect | System design, architecture decisions, ADRs |
code-reviewer | Deep code quality review, security, maintainability |
debugger | Root cause debugging specialist |
refactorer | Code cleanup, restructuring, modernization |
| Agent | Framework | Specialty |
|---|---|---|
nextjs-expert | Next.js 15 | App Router, Server Components, streaming |
react-expert | React 19 | Hooks, performance, concurrent features |
fastapi-expert | FastAPI | Async Python APIs, Pydantic, OpenAPI |
django-expert | Django | ORM, REST framework, admin |
vue-expert | Vue 3 | Composition API, Pinia, Nuxt |
svelte-expert | SvelteKit | Svelte 5, file routing, SSR |
astro-expert | Astro | Islands architecture, content collections |
nestjs-expert | NestJS | TypeScript backend, decorators, DI |
| Agent | Language | Focus |
|---|---|---|
python-expert | Python 3.12+ | Modern patterns, type hints, async |
typescript-expert | TypeScript | Strict mode, advanced types, generics |
javascript-expert | JavaScript | ES2024, patterns, performance |
go-expert | Go | Concurrency, idiomatic patterns, interfaces |
rust-expert | Rust | Ownership, lifetimes, systems programming |
| Agent | What it does |
|---|---|
postgresql-expert | Query optimization, indexes, migrations, partitioning |
mongodb-expert | Aggregations, schema design, Atlas |
redis-expert | Caching strategies, pub/sub, data structures |
database-architect | Database design, normalization, scaling patterns |
| Agent | What it does |
|---|---|
aws-expert | AWS services, IAM, CDK, deployment patterns |
cicd-expert | GitHub Actions, pipelines, deployment automation |
devops-engineer | Docker, containers, observability |
kubernetes-expert | Kubernetes, Helm, service mesh, autoscaling |
terraform-expert | Terraform, IaC, cloud provisioning |
| Agent | What it does |
|---|---|
playwright-expert | Playwright E2E testing, page objects, CI integration |
security-auditor | OWASP Top 10, code security review, threat modeling |
accessibility-expert | WCAG 2.1, screen readers, a11y testing |
e2e-testing-expert | End-to-end test strategy and implementation |
performance-engineer | Core Web Vitals, bundle analysis, rendering perf |
devsecops-engineer | Security in CI/CD, SAST, dependency scanning |
security-pentester | Penetration testing, vulnerability assessment |
| Agent | What it does |
|---|---|
react-native-expert | React Native, Expo, navigation, native modules |
flutter-expert | Flutter, Dart, cross-platform, state management |
| Agent | What it does |
|---|---|
ai-engineer | LLM integration, prompt engineering, RAG, agents |
data-engineer | Data pipelines, ETL, warehouses, Spark |
ml-engineer | ML models, training pipelines, deployment |
bi-expert | Business intelligence, dashboards, SQL analytics |
| Agent | What it does |
|---|---|
technical-writer-dev | Technical documentation for developers |
api-documenter | OpenAPI, API reference docs |
code-documenter | JSDoc, docstrings, inline documentation |
prd-writer | Product requirements documents |
tech-lead | Technical leadership, architecture decisions |
project-manager | Project planning, Agile, delivery |
| Agent | What it does |
|---|---|
ui-components-expert | Component library design, Storybook, design systems |
ux-designer | UX flows, user journeys, wireframe direction |
seo-expert | SEO implementation, structured data, Core Web Vitals |
payment-expert | Stripe, payment flows, PCI compliance |
graphql-expert | GraphQL schema, resolvers, Apollo, Federation |
websocket-expert | WebSockets, real-time systems, Socket.io |
ai-writing-auditor | Detect and eliminate AI-sounding writing patterns |
🎨 Built-in Design System
The kit ships with a production-grade design system — activated with /design. No configuration needed.
7 visual styles:
| Style | Aesthetic | Best for |
|---|---|---|
bauhaus (default) | Geometric, 0 border-radius, IBM Plex Mono, beige/charcoal | Dashboards, tools, institutional |
editorial | Magazine, serif body, blue accents | Blogs, content, portfolios |
brutalist | Raw, black/white, heavy borders | Portfolios, art, attention-grabbing |
agency | Premium consultancy, cream/navy/gold | Landing pages, B2B |
glass | Dark, backdrop-blur, translucent cards | SaaS dashboards |
dark | Near-black, developer terminal aesthetic | Dev tools, CLIs |
custom | Describe your own palette and Claude generates tokens | Any brand |
What /design produces for each request:
- Ready-to-paste React + Tailwind v4 component code
- Full CSS token set for the chosen style
- Precise spacing, typography scale, and color usage guide
- WCAG AA compliant by default
/design a pricing page with 3 tiers
/design editorial a blog post layout
/design glass a user settings panel
/design --tokens ← outputs the full CSS custom property set
🔑 API Keys
Works immediately (no key needed)
- Playwright — screenshots, browser automation
- Memory — remembers facts across sessions
Get these first (5 min each, free)
| Service | Get key | Free tier |
|---|---|---|
| Brave Search | api.search.brave.com | 2,000 searches/mo |
| GitHub | github.com/settings/tokens | Unlimited |
Add when you need them
| Service | Get key | What it unlocks |
|---|---|---|
| OpenRouter | openrouter.ai/keys | GPT-4o, Gemini 2.5, Llama, 100+ models |
| Firecrawl | firecrawl.dev | Web scraping, 500 pages/mo free |
| Perplexity | perplexity.ai/settings/api | AI research synthesis |
| Replicate | replicate.com/account/api-tokens | Image/video/audio generation |
| Apify | apify.com | 4,000+ specialized scrapers |
| Google Maps | console.cloud.google.com | Places, geocoding ($200 free credit) |
Where to add them:
Find the Claude Desktop config file and replace each YOUR_*_KEY placeholder:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
🚀 Start Here — First Steps
After install, try these in order:
No key needed:
Take a Playwright screenshot of google.com
→ Proves your setup works. Should complete in seconds.
Plan your first project [BEGINNER]
/prd
→ Claude asks 8 questions and writes a complete requirements doc.
See the design system [BEGINNER]
/design a landing page with a hero, features, and CTA
→ Get production-ready React + Tailwind code, immediately.
Get expert feedback on an idea [BEGINNER]
/plato quick [describe your idea in one sentence]
→ 3 expert advisors stress-test your idea in under 2 minutes.
Build end-to-end [INTERMEDIATE]
/kickoff build a dashboard that shows my GitHub repo stats
→ Claude architects, designs, builds, and reviews. You approve at each gate.
Learn while building [BEGINNER]
/sensei
→ Before every code task, Claude names the concept, asks you to predict the outcome, then teaches and builds.
Advanced:
/loki-mode — autonomous multi-agent build (no gates)
/autopilot — describe it once, Claude ships it
/video-forge — TTS-narrated video (needs ffmpeg + Remotion + voice API)
⚙️ What the Settings Do
The kit writes ~/.claude/settings.json with three active hooks:
| Hook | When it fires | What it does |
|---|---|---|
| PreToolUse | Before any npm install or yarn add | Blocks it and says "use bun" |
| PostToolUse | After editing any .ts, .tsx, .css file | Reminds you to verify the build |
| Stop | When Claude finishes a session | Asks if you took screenshots and ran tests |
Plus permissions pre-approved for common operations (git, bun, npx, curl, gh) so you don't get prompted for every command.
📁 Repo Structure
claude-code-powerkit/
├── BOOTSTRAP.md ← The installer — paste URL into Claude Code
├── README.md
├── setup.sh ← Partial terminal alternative (skills only, no agents)
├── TIPS.md ← 12 production lessons from real usage
├── SKILLS.md ← Full skills catalog with sources
├── MCP-SERVERS.md ← Detailed MCP setup guide + key links
├── REPOS.md ← 40+ repos for skills, agents, MCP servers
│
├── configs/
│ ├── settings.json ← Claude Code settings (hooks + permissions + plugins)
│ ├── mcp.json ← 10 MCP server template (fill in API keys)
│ ├── CLAUDE.md ← Project instructions template
│ └── rules/
│ ├── frontend.md ← React + Tailwind coding rules
│ └── api.md ← API coding rules
│
├── skills/ ← 30 curated skill files
│ ├── sensei.md
│ ├── design.md
│ ├── plato.md
│ ├── video-forge.md
│ └── ... (27 more)
│
├── agents/ ← 52 developer specialist subagents
│ ├── nextjs-expert.md
│ ├── security-auditor.md
│ └── ... (50 more)
│
└── design-system/
├── tokens.css ← 300+ CSS custom properties
└── quick-start.md ← 5-minute integration guide
🔧 Manual Install (Terminal)
⚠️ This installs skills, settings, and CLAUDE.md only — it does not install the 70 subagents or 24 Composio skills. For the full install, use the BOOTSTRAP.md URL method above.
git clone https://github.com/Neens6655/claude-code-powerkit.git
cd claude-code-powerkit
bash setup.sh
Then copy configs/mcp.json to your Claude Desktop config location and fill in API keys.
🔗 More Resources
- TIPS.md — 12 production lessons from real usage
- MCP-SERVERS.md — Detailed setup guide for all 10 servers
- SKILLS.md — Full skills catalog with install sources
- REPOS.md — 40+ repos for more skills, agents, and MCP servers
- Anthropic Claude Code docs
📄 License
MIT. Fork it, modify it, share it.
Last verified: 2026-05-04 · Claude Code v1.x · v1.0.0