Cloudflare Full-Stack Scaffold

Complete, production-ready starter project for building full-stack applications on Cloudflare with React, Hono, AI SDK, and all Cloudflare services pre-configured.

When to Use This Skill

Use this skill when you need to:

• Start a new full-stack Cloudflare project in minutes instead of hours
• Build AI-powered applications with chat interfaces, RAG, or tool calling
• Have core Cloudflare services ready (D1, KV, R2, Workers AI)
• Opt-in to advanced features (Clerk Auth, AI Chat, Queues, Vectorize)
• Use modern best practices (Tailwind v4, shadcn/ui, AI SDK, React 19)
• Include planning docs and session handoff from the start
• Choose your AI provider (Workers AI, OpenAI, Anthropic, Gemini)
• Enable features only when needed with simple npm scripts
• Avoid configuration errors and integration issues

What This Skill Provides

Complete Scaffold Project

A fully working application you can copy, customize, and deploy immediately:

# Copy the scaffold
cp -r scaffold/ my-new-app/
cd my-new-app/

# Install dependencies
npm install

# Initialize core services (D1, KV, R2)
./scripts/init-services.sh

# Create database tables
npm run d1:local

# Start developing
npm run dev

Result: Full-stack app running in ~5 minutes with:

• ✅ Frontend and backend connected
• ✅ Core Cloudflare services configured (D1, KV, R2, Workers AI)
• ✅ AI SDK ready with multiple providers
• ✅ Planning docs and session handoff protocol
• ✅ Dark mode, theming, UI components
• ✅ Optional features (1 script each to enable):
  • Clerk Auth (npm run enable-auth)
  • AI Chat UI (npm run enable-ai-chat)
  • Queues (npm run enable-queues)
  • Vectorize (npm run enable-vectorize)

Scaffold Structure

scaffold/
├── package.json              # All dependencies (React, Hono, AI SDK, Clerk)
├── tsconfig.json            # TypeScript config
├── vite.config.ts           # Cloudflare Vite plugin
├── wrangler.jsonc           # All Cloudflare services configured
├── .dev.vars.example        # Environment variables template
├── .gitignore               # Standard ignores
├── README.md                # Project-specific readme
├── CLAUDE.md                # Project instructions for Claude
├── SCRATCHPAD.md            # Session handoff protocol
├── CHANGELOG.md             # Version history
├── schema.sql               # D1 database schema
│
├── docs/                    # Complete planning docs
│   ├── ARCHITECTURE.md
│   ├── DATABASE_SCHEMA.md
│   ├── API_ENDPOINTS.md
│   ├── IMPLEMENTATION_PHASES.md
│   ├── UI_COMPONENTS.md
│   └── TESTING.md
│
├── migrations/              # D1 migrations
│   └── 0001_initial.sql
│
├── src/                     # Frontend (React + Vite + Tailwind v4)
│   ├── main.tsx
│   ├── App.tsx
│   ├── index.css           # Tailwind v4 theming
│   ├── components/
│   │   ├── ui/             # shadcn/ui components
│   │   ├── ThemeProvider.tsx
│   │   ├── ProtectedRoute.tsx     # Auth (COMMENTED)
│   │   └── ChatInterface.tsx      # AI chat (COMMENTED)
│   ├── lib/
│   │   ├── utils.ts        # cn() utility
│   │   └── api-client.ts   # Fetch wrapper
│   └── pages/
│       ├── Home.tsx
│       ├── Dashboard.tsx
│       └── Chat.tsx        # AI chat page (COMMENTED)
│
└── backend/                 # Backend (Hono + Cloudflare)
    ├── src/
    │   └── index.ts        # Main Worker entry
    ├── middleware/
    │   ├── cors.ts
    │   └── auth.ts         # JWT (COMMENTED)
    ├── routes/
    │   ├── api.ts          # Basic API routes
    │   ├── d1.ts           # D1 examples
    │   ├── kv.ts           # KV examples
    │   ├── r2.ts           # R2 examples
    │   ├── ai.ts           # Workers AI (direct binding)
    │   ├── ai-sdk.ts       # AI SDK examples (multiple providers)
    │   ├── vectorize.ts    # Vectorize examples
    │   └── queues.ts       # Queues examples
    └── db/
        └── queries.ts      # D1 typed query helpers

Helper Scripts

scripts/setup-project.sh:

• Copies scaffold to new directory
• Renames project in package.json
• Initializes git repository
• Runs npm install
• Prompts to initialize services

scripts/init-services.sh:

• Creates D1 database via wrangler
• Creates KV namespace
• Creates R2 bucket
• Updates wrangler.jsonc with IDs
• (Queues and Vectorize created when enabled)

scripts/enable-auth.sh:

• Uncomments all Clerk auth patterns
• Enables ProtectedRoute component
• Enables auth middleware
• Prompts for Clerk API keys
• Updates .dev.vars

scripts/enable-ai-chat.sh:

• Uncomments ChatInterface component
• Uncomments Chat page
• Enables AI SDK UI patterns
• Adds chat route to App.tsx
• Prompts for AI provider API keys

scripts/enable-queues.sh:

• Uncomments Queues routes and bindings
• Enables async message processing
• Provides queue creation instructions
• Updates backend and config files

scripts/enable-vectorize.sh:

• Uncomments Vectorize routes and bindings
• Enables vector search and RAG
• Provides index creation instructions
• Configures embedding dimensions

Reference Documentation

references/quick-start-guide.md:

• 5-minute setup walkthrough
• First deployment guide
• Common customizations

references/service-configuration.md:

• Details on each Cloudflare service
• When to use each one
• Configuration options

references/ai-sdk-guide.md:

• AI SDK Core vs UI
• Provider switching patterns
• Streaming and tool calling
• RAG implementation

references/customization-guide.md:

• Removing unused services
• Adding new routes/pages
• Customizing theme
• Project structure best practices

references/enabling-auth.md:

• Clerk setup walkthrough
• JWT template configuration
• Testing auth flow

Key Integrations

1. AI SDK Integration (Three Approaches)

Direct Workers AI Binding (fastest):

// Already works, no API key needed
const result = await c.env.AI.run('@cf/meta/llama-3-8b-instruct', {
  messages: [{ role: 'user', content: 'Hello' }]
})

AI SDK with Workers AI (portable code, same infrastructure):

import { streamText } from 'ai'
import { createWorkersAI } from 'workers-ai-provider'

const workersai = createWorkersAI({ binding: c.env.AI })
const result = await streamText({
  model: workersai('@cf/meta/llama-3-8b-instruct'),
  messages: [{ role: 'user', content: 'Hello' }]
})

AI SDK with External Providers (OpenAI, Anthropic, Gemini):

import { openai } from '@ai-sdk/openai'
import { anthropic } from '@ai-sdk/anthropic'

// Switch providers in 1 line
const result = await streamText({
  model: openai('gpt-4o'),  // or anthropic('claude-sonnet-4-5')
  messages: [{ role: 'user', content: 'Hello' }]
})

AI SDK v5 UI - Chat Interface (COMMENTED, enable with script):

import { useChat } from '@ai-sdk/react'
import { DefaultChatTransport } from 'ai'
import { useState } from 'react'

function ChatInterface() {
  const [input, setInput] = useState('')
  const { messages, sendMessage, status } = useChat({
    transport: new DefaultChatTransport({
      api: '/api/ai-sdk/chat',
    }),
  })

  // Send message on Enter key
  const handleKeyDown = (e) => {
    if (e.key === 'Enter' && status === 'ready' && input.trim()) {
      sendMessage({ text: input })
      setInput('')
    }
  }

  // Render messages (v5 uses message.parts[])
  return (
    <div>
      {messages.map(m => (
        <div key={m.id}>
          {m.parts.map(part => {
            if (part.type === 'text') return <div>{part.text}</div>
          })}
        </div>
      ))}
      <input
        value={input}
        onChange={e => setInput(e.target.value)}
        onKeyDown={handleKeyDown}
        disabled={status !== 'ready'}
      />
    </div>
  )
}

2. Forms & Data Fetching (React Hook Form + Zod + TanStack Query)

Industry-Standard Libraries for Production Apps:

React Hook Form - Performant form state management:

import { useForm } from 'react-hook-form'
import { zodResolver } from '@hookform/resolvers/zod'

const form = useForm({
  resolver: zodResolver(userSchema), // Zod validation
})

<input {...register('name')} />
{errors.name && <span>{errors.name.message}</span>}

Zod v4 - TypeScript-first schema validation:

// Define schema once
const userSchema = z.object({
  name: z.string().min(2),
  email: z.string().email(),
  age: z.number().int().positive().optional(),
})

// Infer TypeScript type
type User = z.infer<typeof userSchema>

// Use in frontend (React Hook Form)
resolver: zodResolver(userSchema)

// Use in backend (same schema!)
const validated = userSchema.parse(requestBody)

TanStack Query v5 - Smart data fetching & caching:

// Fetch data with automatic caching
const { data, isLoading } = useQuery({
  queryKey: ['users'],
  queryFn: () => apiClient.get('/api/users'),
})

// Update data with mutations
const mutation = useMutation({
  mutationFn: (newUser) => apiClient.post('/api/users', newUser),
  onSuccess: () => {
    queryClient.invalidateQueries({ queryKey: ['users'] })
  },
})

Full-Stack Validation Pattern:

• ✅ Define schema in shared/schemas/ (single source of truth)
• ✅ Frontend validates instantly (React Hook Form + Zod)
• ✅ Backend validates securely (same Zod schema)
• ✅ TypeScript types inferred automatically
• ✅ Update validation once, applies everywhere

Complete Working Examples:

• Profile page with form: /profile route
• Dashboard with queries: /dashboard route
• Form component: src/components/UserProfileForm.tsx
• Backend validation: backend/routes/forms.ts
• Shared schemas: shared/schemas/userSchema.ts

See references/supporting-libraries-guide.md for comprehensive guide.

3. All Cloudflare Services Pre-Configured

Database (D1):

• Schema file with example tables
• Migrations directory
• Typed query helpers
• Example CRUD routes

Key-Value Storage (KV):

• Get/put/delete examples
• TTL patterns
• Bulk operations

Object Storage (R2):

• Upload/download examples
• Presigned URLs
• Streaming large files

AI Inference (Workers AI):

• Text generation
• Embeddings
• Image generation (Stable Diffusion)

Vector Database (Vectorize):

• Insert/query embeddings
• RAG patterns
• Similarity search

Message Queues:

• Producer examples
• Consumer patterns
• Batch processing

3. Optional Clerk Authentication

All auth patterns included but COMMENTED - uncomment to enable:

./scripts/enable-auth.sh
# Prompts for Clerk keys, uncomments all patterns

What gets enabled:

• Frontend: ProtectedRoute component, auth in api-client
• Backend: JWT verification middleware
• Protected API routes
• Auth loading states
• Session management

4. Planning Docs + Session Handoff Protocol

docs/ directory - Complete planning structure:

• ARCHITECTURE.md: System design
• DATABASE_SCHEMA.md: D1 schema docs
• API_ENDPOINTS.md: All routes documented
• IMPLEMENTATION_PHASES.md: Phased build approach
• UI_COMPONENTS.md: Component hierarchy
• TESTING.md: Test strategy

SCRATCHPAD.md - Session handoff protocol:

• Current phase tracking
• Progress checkpoints
• Next actions
• References to planning docs

Usage Guide

Quick Start (5 Minutes)

# 1. Copy scaffold
cd /path/to/skills/cloudflare-full-stack-scaffold
cp -r scaffold/ ~/projects/my-new-app/
cd ~/projects/my-new-app/

# 2. Run setup
npm install

# 3. Initialize Cloudflare services
npx wrangler d1 create my-app-db
npx wrangler kv:namespace create my-app-kv
npx wrangler r2 bucket create my-app-bucket
npx wrangler vectorize create my-app-index --dimensions=1536
npx wrangler queues create my-app-queue

# 4. Update wrangler.jsonc with IDs from step 3

# 5. Create D1 tables
npx wrangler d1 execute my-app-db --local --file=schema.sql

# 6. Start dev server
npm run dev

Visit: http://localhost:5173

Enable Authentication (Optional)

./scripts/enable-auth.sh
# Prompts for Clerk publishable and secret keys
# Uncomments all auth patterns
# Updates .dev.vars

npm run dev

Enable AI Chat Interface (Optional)

./scripts/enable-ai-chat.sh
# Uncomments ChatInterface component
# Uncomments Chat page
# Prompts for OpenAI/Anthropic API keys (optional)

npm run dev

Visit: http://localhost:5173/chat

Deploy to Production

# Build
npm run build

# Deploy
npx wrangler deploy

# Migrate production database
npx wrangler d1 execute my-app-db --remote --file=schema.sql

# Set production secrets
npx wrangler secret put CLERK_SECRET_KEY
npx wrangler secret put OPENAI_API_KEY

Customization Patterns

Remove Unused Services

Don't need Vectorize?

1. Delete backend/routes/vectorize.ts
2. Remove vectorize binding from wrangler.jsonc
3. Remove from vite.config.ts cloudflare plugin
4. Remove route registration in backend/src/index.ts

Add New API Routes

// backend/routes/my-feature.ts
import { Hono } from 'hono'

export const myFeatureRoutes = new Hono()

myFeatureRoutes.get('/hello', (c) => {
  return c.json({ message: 'Hello from my feature!' })
})

// backend/src/index.ts
import { myFeatureRoutes } from './routes/my-feature'
app.route('/api/my-feature', myFeatureRoutes)

Switch AI Providers

// Change this line:
model: openai('gpt-4o'),

// To this:
model: anthropic('claude-sonnet-4-5'),

// Or this:
model: google('gemini-2.5-flash'),

// Or use Workers AI:
const workersai = createWorkersAI({ binding: c.env.AI })
model: workersai('@cf/meta/llama-3-8b-instruct'),

Customize Theme

All theming in src/index.css:

:root {
  --background: hsl(0 0% 100%);  /* Change colors here */
  --foreground: hsl(0 0% 3.9%);
  --primary: hsl(220 90% 56%);
  /* etc */
}

Architecture Highlights

Frontend-Backend Connection

Key Insight: The Vite plugin runs your Worker on the SAME port as Vite.

// ✅ CORRECT: Use relative URLs
fetch('/api/data')

// ❌ WRONG: Don't use absolute URLs
fetch('http://localhost:8787/api/data')

No proxy configuration needed!

Environment Variables

Frontend (.env):

VITE_CLERK_PUBLISHABLE_KEY=pk_test_xxx

Backend (.dev.vars):

CLERK_SECRET_KEY=sk_test_xxx
OPENAI_API_KEY=sk-xxx

CORS Configuration

Critical: CORS middleware must be applied BEFORE routes:

// ✅ CORRECT ORDER
app.use('/api/*', corsMiddleware)
app.post('/api/data', handler)

// ❌ WRONG - Will cause CORS errors
app.post('/api/data', handler)
app.use('/api/*', corsMiddleware)

Auth Pattern (When Enabled)

Frontend: Check isLoaded before making API calls:

const { isLoaded, isSignedIn } = useSession()

useEffect(() => {
  if (!isLoaded) return  // Wait for auth
  fetch('/api/protected').then(/* ... */)
}, [isLoaded])

Backend: JWT verification middleware:

import { jwtAuthMiddleware } from './middleware/auth'

app.use('/api/protected/*', jwtAuthMiddleware)

Dependencies Included

{
  "dependencies": {
    "react": "^19.2.0",
    "react-dom": "^19.2.0",
    "hono": "^4.10.2",
    "@cloudflare/vite-plugin": "^1.13.14",

    "ai": "^5.0.76",
    "@ai-sdk/openai": "^1.0.0",
    "@ai-sdk/anthropic": "^1.0.0",
    "@ai-sdk/google": "^1.0.0",
    "workers-ai-provider": "^2.0.0",
    "@ai-sdk/react": "^1.0.0",

    "@clerk/clerk-react": "^5.53.3",
    "@clerk/backend": "^2.19.0",

    "tailwindcss": "^4.1.14",
    "@tailwindcss/vite": "^4.1.14",
    "class-variance-authority": "^0.7.0",
    "clsx": "^2.1.1",
    "tailwind-merge": "^2.5.4",

    "zod": "^3.24.1",
    "react-hook-form": "^7.54.2",
    "@hookform/resolvers": "^3.9.1",

    "@tanstack/react-query": "^5.62.11",

    "@radix-ui/react-slot": "^1.1.1",
    "@radix-ui/react-dropdown-menu": "^2.1.4"
  },
  "devDependencies": {
    "@types/react": "^19.0.0",
    "@types/react-dom": "^19.0.0",
    "typescript": "^5.7.2",
    "vite": "^7.1.11",
    "wrangler": "^4.0.0"
  }
}

Token Efficiency

Scenario	Without Scaffold	With Scaffold	Savings
Initial setup	~18-22k tokens	~3-5k tokens	~80%
Service configuration	~8-10k tokens	0 tokens (done)	100%
Frontend-backend connection	~5-7k tokens	0 tokens (done)	100%
AI SDK setup	~4-6k tokens	0 tokens (done)	100%
Auth integration	~6-8k tokens	~500 tokens	~90%
Planning docs	~3-5k tokens	0 tokens (included)	100%
Total	~44-58k tokens	~3-6k tokens	~90%

Time Savings: 3-4 hours → 5-10 minutes (~95% faster)

Common Issues Prevented

Issue	How Scaffold Prevents It
Service binding errors	All bindings pre-configured and tested
CORS errors	Middleware in correct order
Auth race conditions	Proper loading state patterns
Frontend-backend connection	Vite plugin correctly configured
AI SDK setup confusion	Multiple working examples
Missing planning docs	Complete docs/ structure included
Environment variable mix-ups	Clear .dev.vars.example with comments
Missing migrations	migrations/ directory with examples
Inconsistent file structure	Standard, tested structure
Database type errors	Typed query helpers included
Theme configuration	Tailwind v4 theming pre-configured
Build errors	Working build config (vite + wrangler)

Total Errors Prevented: 12+ common setup and integration errors

When NOT to Use This Scaffold

• ❌ Building a static site (no backend needed)
• ❌ Using Next.js, Remix, or other meta-framework
• ❌ Need SSR (use framework-specific Cloudflare adapter)
• ❌ Building backend-only API (no frontend needed)
• ❌ Extremely simple single-page app

For these cases: Use minimal templates or official framework starters.

Production Evidence

Based on:

• Cloudflare's official agents-starter template (AI SDK patterns)
• cloudflare-full-stack-integration skill (tested frontend-backend patterns)
• session-handoff-protocol skill (planning docs + SCRATCHPAD.md)
• tailwind-v4-shadcn skill (UI component patterns)
• Multiple production Jezweb projects

Package versions verified: 2025-10-23

Works with:

• Cloudflare Workers (production environment)
• Wrangler 4.0+
• Node.js 18+
• npm/pnpm/yarn

Quick Reference

Setup new project:

cp -r scaffold/ my-app/
cd my-app/
npm install
# Follow quick-start-guide.md

Enable auth:

./scripts/enable-auth.sh

Enable AI chat:

./scripts/enable-ai-chat.sh

Deploy:

npm run build
npx wrangler deploy

Key Files:

• wrangler.jsonc - Service configuration
• vite.config.ts - Build configuration
• .dev.vars.example - Environment variables template
• docs/ARCHITECTURE.md - System design
• SCRATCHPAD.md - Session handoff protocol

---

Remember: This scaffold is a starting point, not a constraint. Customize everything to match your needs. The value is in having a working foundation with all the integration patterns already figured out, saving hours of setup and debugging time.