Transform ideas into production-ready applications through structured AI-led development
MegaPrompt V8 is a comprehensive framework that enables AI models (Claude, GPT, Gemini) to transform natural language descriptions into complete, production-ready applications through structured phases and intelligent execution.
Key Philosophy:
- π§ AI-Led Execution: AI drives technical implementation
- π€ User-Led Vision: Users define goals and requirements
- π€ Collaborative Partnership: Continuous dialogue and iteration
- π Evidence-Based Decisions: All recommendations backed by data
- β‘ Proactive Intelligence: AI suggests next steps autonomously
- β Phase-Based Development: Structured flow from discovery to delivery
- β Multi-Model Support: Works with Claude, GPT, Gemini
- β Comprehensive Documentation: 30+ files auto-generated
- β Cost Optimization: Real-time tracking and multi-model strategies
- β Security First: Automated security checks before delivery
- β Quality Gates: 80%+ test coverage, automated validation
- β Technology Radar: Continuous evaluation of new technologies
- β Scalability Planning: Plan for scale, build essentials now
- π Dynamic Role Switching: 30+ expert roles (DevOps, Security, ML, etc.)
- π¨ Flexible UI/UX Modes: Support for templates, custom designs, AI generation
- πΎ Context Memory: Persistent decision tracking across sessions
- π¨ Telegram Notifications: Real-time progress alerts
- π Sound Alerts: Audio notifications for key events
- π¦ Git Auto-Backup: Automated version control every 5 tasks
- π€ Multi-Agent Coordination: Parallel execution with multiple AI models
1. Identify Your Model
β‘ Claude β Read CLAUDE.md
β‘ GPT β Read CODEX.md
β‘ Gemini β Read GEMINI.md
2. Load Framework
Read: MEGAPROMPT_V8_FINAL.md (complete instructions)
Read: AGENT_START_PROMPT.md (quick setup)
3. Start Working
Begin with Phase 0 greeting
Follow structured workflow
Generate comprehensive documentation
1. Choose Your AI
- Claude Sonnet 4.5 (balanced, fast)
- Claude Opus 4.1 (strongest reasoning)
- GPT-4o (UI/UX excellence)
- Gemini 2.0 Flash (speed, efficiency)
2. Start Conversation
User: "I want to build [your idea]"
AI: [Follows MegaPrompt V8 workflow]
3. Collaborate
- AI asks questions (Discovery)
- You provide answers
- AI suggests options
- You make decisions
- AI builds everything
4. Receive Deliverable
- Complete documentation
- Production-ready code
- Tests (80%+ coverage)
- README and deployment guide
- ZIP package ready to use
MEGAPROMPT_V8_FINAL.md # Core framework (complete instructions)
AGENT_START_PROMPT.md # Quick start guide
CLAUDE.md # Claude-specific instructions
CODEX.md # GPT-specific instructions
GEMINI.md # Gemini-specific instructions
.cursorrules # Cursor IDE integration
README.md # This file
docs/
00_context/ # Requirements & decisions
PRD.md # User's vision (their words)
PRD_ai.md # Executable plan
PRD_NOTES.md # Execution log
PRD_IMPLEMENTATION_MATRIX.md # FeatureβTaskβFile mapping (if >20 tasks)
DECISIONS.md # All technical decisions
GLOSSARY.md # Domain terminology
CONTEXT_MEMORY.md # Persistent memory
10_product/ # Features & planning
SPEC.md # Detailed specifications
TASKS.md # Task breakdown
MVP_CHECKLIST.md # Scope control
20_engineering/ # Architecture & tech
ARCHITECTURE.md # System design
RULES.md # Coding standards
TECH_STACK.md # Technology rationale
DATABASE_SCHEMA.md # Data model
SECURITY_CHECKLIST.md # Security protocol
30_design/ # UI/UX decisions
UIUX.md # Design decisions
UI_PAGES.md # Page inventory
UI_PROMPTS.md # AI generation prompts
UI_MODE.md # Design workflow
40_api/ # API specifications
OPENAPI.yaml # API spec (OpenAPI 3.0)
OPENSPEC_WORKFLOW.md # API workflow
50_testing/ # Testing strategy
TEST_STRATEGY.md # Testing approach
90_ops/ # Operations & tracking
CHANGELOG.md # Change history
PROGRESS.md # Real-time status
COST_TRACKING.md # Budget dashboard
TIMELINE_AND_COST.md # Planning
DEPLOYMENT_GUIDE.md # Deployment steps
SCALE_PLAN.md # Growth strategy
TELEGRAM_SETUP.md # Notification config
SOUND_NOTIFICATION_SETUP.md # Alert config
GIT_AUTO_BACKUP.md # Version control
src/ # Source code
tests/ # Test files
README.md # Project readme
.env.example # Environment template
package.json # Dependencies
AI greets user in their language
Explains the process
Asks what they want to build
AI asks strategic questions:
- What problem does this solve?
- Who are the users?
- What are the core features?
- Any constraints (time, budget, scale)?
AI extracts:
- User personas
- Use cases
- Data entities
- Scale requirements
AI proposes:
- Complete feature list (MVP vs Later vs Never)
- Technology options (with pros/cons/evidence)
- Architecture approach
- Timeline estimate
- Cost projection
User approves each decision
AI documents everything
Triggered by: "BUILD" or "Ψ¨Ψ³Ψ§Ψ²"
AI generates:
- All documentation (30+ files)
- Complete source code
- Configuration files
- Tests (80%+ coverage)
- README and guides
AI updates PROGRESS.md after each task
AI suggests next steps proactively
AI performs:
- Security scan (auto-check)
- Quality gates validation
- Test coverage verification
- Documentation completeness
- Final README generation
AI packages:
- ZIP file with everything
- Clear installation instructions
- Quick start guide
For projects with 20+ tasks, MegaPrompt V8 can coordinate multiple AI models:
High Complexity (Architecture, Security):
β Claude Opus 4.1 / GPT o1
Medium Complexity (Backend, API):
β Claude Sonnet 4.5 / GPT-4o
Low Complexity (Frontend, UI):
β Gemini Flash 2.0 / GPT-4o
Testing (Fast iteration):
β Gemini Flash 2.0
Documentation:
β Gemini Pro 2.0 / GPT-4o
| Strategy | Duration | Cost | Quality |
|---|---|---|---|
| Single Model | 12h | $25 | β β β β β |
| Multi-Model Balanced | 7h | $28 | β β β β β |
| Multi-Model Speed | 5h | $35 | β β β β β |
| Multi-Model Budget | 10h | $18 | β β β ββ |
User: "I want to build a task management app"
AI: "Great! Let me understand your vision.
What problem does this solve for users?
Who will use it (personal, teams, both)?
Any specific features you need?
Timeline or budget constraints?"
User: [Answers questions]
AI: "Based on your input, here are all features:
MVP (Build Now):
β User authentication
β Create/edit/delete tasks
β Mark tasks as complete
Future (Later):
β Team collaboration
β File attachments
β Email notifications
Out of Scope:
π« Mobile app (use PWA instead)
Approve?"
User: "Yes"
AI: "Technology choices:
Frontend: React vs Vue vs Svelte
[Shows pros/cons/evidence for each]
My recommendation: React
Reason: [specific rationale]
Your choice?"
User: "React"
AI: "Perfect! Ready to build?
Timeline: 2 weeks
Cost: $25 (using Claude Sonnet)
Deliverable: Complete app + docs
Say 'BUILD' when ready"
User: "BUILD"
AI: [Generates all files, provides progress updates]
AI: "π¦ Project complete!
[Download ZIP - 2.3MB]
Quick start:
1. Extract ZIP
2. npm install
3. npm run dev
Enjoy! π"
Before every delivery:
β
API keys in .env (not hardcoded)
β
Passwords hashed (bcrypt/argon2)
β
SQL injection prevention
β
CORS properly configured
β
Rate limiting implemented
β
Dependencies scanned for vulnerabilities
Generated: docs/20_engineering/SECURITY_CHECKLIST.md
Contains:
- Authentication best practices
- Authorization patterns
- Input validation rules
- Error handling guidelines
- Dependency security
- Deployment security
File: docs/90_ops/COST_TRACKING.md
βββββββββββββββββββββββββββββββββββββββββββββββ
β Cost Dashboard β
βββββββββββββββββββββββββββββββββββββββββββββββ€
β Current: $18.50 / $25.00 (74%) β
β Remaining: $6.50 β
β ββββββββββ 74% β
β β
β Projection: $24.00 (On track β) β
β Buffer: $1.00 β
βββββββββββββββββββββββββββββββββββββββββββββββ
Per Model:
- Claude Sonnet: $8.70
- GPT-4o: $6.50
- Gemini Flash: $3.30
Status: β On Track
- Minimum Coverage: 80%
- Unit Tests: All critical functions
- Integration Tests: API endpoints
- E2E Tests: Core user flows
- TypeScript: Type safety enforced
- ESLint: No warnings allowed
- Prettier: Consistent formatting
- Documentation: JSDoc for all functions
β‘ All tasks completed (100%)
β‘ Tests passing (80%+ coverage)
β‘ Security scan clean (no critical issues)
β‘ Dependencies updated (no vulnerabilities)
β‘ Documentation complete (all files)
β‘ README generated (installation guide)
β‘ .env.example provided (configuration)
β‘ Deployment guide ready (step-by-step)
AI continuously evaluates new technologies and suggests when beneficial:
"π‘ New technology available: Bun 1.0
Benefits for this project:
- 3x faster than Node.js
- Built-in TypeScript support
- Drop-in replacement
Would you like to use it?
Migration effort: 30 minutes"
AI remembers decisions across sessions:
File: docs/00_context/CONTEXT_MEMORY.md
Previous decisions:
- Database: PostgreSQL (chosen over MongoDB)
- Auth: JWT (chosen over OAuth initially)
- Styling: Tailwind CSS (chosen over Material UI)
Reasoning preserved for future reference
AI plans for scale but builds essentials:
File: docs/90_ops/SCALE_PLAN.md
Now (MVP):
- Single server
- SQLite database
- Simple caching
1K-10K users:
- Load balancer
- PostgreSQL
- Redis cache
10K+ users:
- Kubernetes
- Database sharding
- CDN
MegaPrompt V8 draws inspiration from:
GitHub: cloudflare/vibesdk
Learnings:
- Phase-wise code generation
- Container-based execution
- Real-time preview capabilities
- Durable Objects for state management
GitHub: cloudflare/agents
Learnings:
- Stateful AI agent patterns
- WebSocket communication
- Persistent memory systems
- Multi-agent coordination
MEGAPROMPT_V8_FINAL.md- Complete framework specificationAGENT_START_PROMPT.md- Quick start for AI agentsCLAUDE.md- Claude-specific instructionsCODEX.md- GPT-specific instructionsGEMINI.md- Gemini-specific instructions
.cursorrules- Cursor IDE configurationREADME.md- This file (GitHub documentation)
Want to improve MegaPrompt V8?
- Fork the repository
- Create your feature branch
- Make improvements
- Test thoroughly
- Submit pull request
Areas for contribution:
- Additional AI model support
- New documentation templates
- Enhanced security checks
- Performance optimizations
- Language translations
MIT License - See LICENSE for details
- Cloudflare team for VibeSDK and Agents frameworks
- AI research community for foundational work
- Open source contributors
- π Documentation
- π¬ Discussions
- π Issues
- Choose your AI model (Claude/GPT/Gemini)
- Load the appropriate guide (CLAUDE.md/CODEX.md/GEMINI.md)
- Read MEGAPROMPT_V8_FINAL.md
- Start building!
Remember: AI leads execution, you lead vision. Together, you create amazing products.
Made with β€οΈ using AI + Human Collaboration
Version 8.0 - Last Updated: 2024-12-31