Build an MVP in hours, not months — guided by AI coding agents
Transform any app idea into working code through 5 AI-powered stages:
flowchart LR
subgraph Phase1[" "]
A[💡 Idea]
end
subgraph Phase2["Research"]
B[📊 Market Analysis]
end
subgraph Phase3["Define"]
C[📋 PRD]
end
subgraph Phase4["Design"]
D[🏗️ Tech Design]
end
subgraph Phase5["Generate"]
E[🤖 AGENTS.md]
end
subgraph Phase6["Build"]
F[🚀 MVP]
end
A --> B --> C --> D --> E --> F
style A fill:#667eea,stroke:#667eea,color:#fff
style B fill:#764ba2,stroke:#764ba2,color:#fff
style C fill:#f093fb,stroke:#f093fb,color:#fff
style D fill:#4facfe,stroke:#4facfe,color:#fff
style E fill:#00f2fe,stroke:#00f2fe,color:#000
style F fill:#43e97b,stroke:#43e97b,color:#000
| Stage | What Happens | Output | Time |
|---|---|---|---|
 |
Validate market & tech landscape | research.txt |
20 min |
 |
Clarify product scope | PRD.md |
15 min |
 |
Decide how to build | TechDesign.md |
15 min |
 |
Convert docs into agent blueprints | AGENTS.md |
10 min |
 |
Generate & test code | Working MVP | 1-3 hrs |
┌─────────────────────────────────────────────────────────────────┐
│ 1. Copy prompt file → 2. Answer questions → 3. Get docs │
│ 4. Feed docs to AI agent → 5. Ship your MVP │
└─────────────────────────────────────────────────────────────────┘
| Step | What You Do | Result |
|---|---|---|
| 1 | Copy part1-deepresearch.md, answer questions |
Research doc |
| 2 | Copy part2-prd-mvp.md, answer questions |
PRD doc |
| 3 | Copy part3-tech-design-mvp.md, answer questions |
Tech Design doc |
| 4 | Copy part4-notes-for-agent.md, generate configs |
AGENTS.md + configs |
| 5 | Tell your AI agent: "Read AGENTS.md and build the MVP" | Working MVP |
Automated alternative: Try the Vibe-Coding Webapp to skip the manual copy-pasting.
For the research and planning phases (Parts 1-4):
| Platform | Best For | Link |
|---|---|---|
 |
Technical accuracy and reasoning | claude.ai |
 |
Large context for comprehensive research | aistudio.google.com |
 |
Iterative research | chat.openai.com |
For the build phase (Part 5):
|
Terminal-Based
|
IDE-Based
|
No-Code
|
- Any modern browser
- 2-4 hours of time
- Basic computer skills (no coding required)
- Optional: Node.js for terminal-based tools
Validate your idea with AI-powered market research — 20-30 min
What this does: Analyzes market opportunity, competitors, and technical feasibility.
How it works:
- Copy the entire
part1-deepresearch.mdfile - Paste into your chosen AI platform (Claude, Gemini, or ChatGPT)
- Answer 5-6 questions tailored to your experience level
- AI generates comprehensive research with market analysis, competitor breakdown, technical recommendations, and cost estimates
- Save output as
research-[YourAppName].txt
Tip: If your platform supports web search, enable it for up-to-date stats and competitor info.
Define exactly what you're building — 15-20 min
What this does: Transforms your idea into clear, actionable product specifications.
How it works:
- Copy
part2-prd-mvp.mdinto a new AI chat - Attach your research findings when prompted
- Answer questions about core features, target users, success metrics, and UI/UX vision
- AI creates professional PRD document
- Save as
PRD-[YourAppName]-MVP.md
Plan the technical architecture — 15-20 min
What this does: Decides the best tech stack and implementation approach.
How it works:
- Copy
part3-tech-design-mvp.mdinto a new AI chat - Attach your PRD (required) and research (optional)
- Answer questions about platform, complexity tolerance, budget, and timeline
- AI recommends optimal stack (no-code, low-code, or full-code)
- Save as
TechDesign-[YourAppName]-MVP.md
Create blueprints for your AI coding assistant — 5-10 min
What this does: Converts all docs into step-by-step coding instructions for AI agents.
How it works:
- Copy
part4-notes-for-agent.mdinto a new AI chat - Attach PRD and Technical Design documents
- Ask the AI for a brief plan for the AGENTS structure; review and approve
- AI generates:
AGENTS.md— Universal instructions- Tool-specific configs based on your choice (CLAUDE.md, .cursorrules, etc.)
- Save all files in your project root
Note: Treat AGENTS.md and tool configs as living docs — update them as your project scales.
Let AI build your MVP — 1-3 hrs
Terminal Agents (Claude Code, Gemini CLI)
cd your-project
claude "Read AGENTS.md and build the MVP"
# or
gemini "Read AGENTS.md and implement"IDE Tools (Cursor, VS Code)
- Open your project folder in the IDE
- Add configuration file (.cursorrules or similar)
- Start with: "Read AGENTS.md and build the MVP step by step"
No-Code Platforms (Lovable, v0)
- Go to platform website
- Paste your PRD content
- Say: "Build this MVP following the specifications"
┌──────────┐ ┌──────────┐ ┌──────────┐
│ Plan │ --> │ Execute │ --> │ Verify │
│ (approve)│ │(one feat)│ │ (test) │
└──────────┘ └──────────┘ └──────────┘
^ │
└─────────────────────────────────┘
| Level | First Prompt |
|---|---|
| Vibe-coder | "Read AGENTS.md and agent_docs. Propose a plan first, wait for approval, then build step by step." |
| Developer | "Review AGENTS.md and architecture. Propose a Phase 1 plan, get approval, then implement with proper patterns." |
Follow-up prompts:
- "What's done and what's left?"
- "Test [feature] and fix any issues"
- "Make it work on mobile"
- "Deploy to Vercel/Cloudflare"
your-app/
├── 📁 docs/
│ ├── research-YourApp.txt
│ ├── PRD-YourApp-MVP.md
│ └── TechDesign-YourApp-MVP.md
├── 📁 agent_docs/
│ ├── tech_stack.md
│ ├── code_patterns.md
│ ├── project_brief.md
│ ├── product_requirements.md
│ └── testing.md
├── 📄 AGENTS.md # Universal AI instructions
├── 📄 CLAUDE.md # Claude Code config (if using)
├── 📄 .cursorrules # Cursor config (if using)
└── 📁 src/ # Your application code
Once your MVP is built, deploy to one of these platforms:
| Platform | Best For | Free Tier |
|---|---|---|
 |
Next.js, React, frontend apps | ✓ |
 |
Static sites, edge functions | ✓ |
Both platforms support git-based deployments — push your code and it deploys automatically.
Which tools should I use?
| Situation | Recommended Tools |
|---|---|
| Complete beginner |  or  → instant hosted apps |
| Learning to code |  or VS Code with Copilot |
| Experienced developer |  or Cursor |
| Budget-limited |  (free) + VS Code |
| Need MVP fast | Lovable → quick prototypes |
| Complex logic | Claude Code → session memory for large codebases |
When NOT to use these tools:
- Native mobile or hardware builds — use traditional toolchains
- Regulated workloads (SOC2, HIPAA) — use enterprise solutions
- Safety-critical systems — require deterministic, human-led engineering
Avoid these mistakes
| Pitfall | Solution |
|---|---|
| Skipping discovery work | Run the Part 1 research prompt first |
| Letting agents ship code alone | Review the diff and run tests before merging |
| Publishing auto-generated UIs without checks | Test accessibility and security before launch |
| Forcing one tool to do everything | Mix tools — IDE + terminal + builder |
Quick fixes for common issues
| Problem | Solution |
|---|---|
| "AI ignores my documents" | Start with: "First read AGENTS.md, PRD, and TechDesign. Summarize key requirements before coding." |
| "Code doesn't match PRD" | Say: "Re-read the PRD section on [feature], list acceptance criteria, then refactor accordingly." |
| "AI is overcomplicating" | Add to config: "Prioritize MVP scope. Offer the simplest working implementation." |
| "Deployment failing" | Request: "Walk through deployment checklist, verify env vars, then run health check." |
PRs and issues welcome! Help us improve:
- Report issues with prompts
- Share your success stories
- Add new tool configurations
- Submit example MVPs built with this workflow
See CONTRIBUTING.md for guidelines.
Released under the MIT License.
The best time to build your idea was yesterday. The second best time is now.
Created by the vibe-coding community
