Conversational MVP Builder

The MVP Builder is Autoflowly's core feature. You describe a business idea in plain English, and the platform generates a complete, deployed full-stack application.

How It Works

The generation pipeline has four stages:

1. Idea Analysis

When you submit a description, the system:

Extracts the industry, target audience, and core features
Identifies the best-fit template from 54+ prebuilt options
Determines the tech stack (React/Next.js frontend, FastAPI backend, PostgreSQL)

2. Template-AI Hybrid Generation

Autoflowly uses a hybrid architecture — combining prebuilt templates with AI code generation:

Component	How It's Built
Project structure	Prebuilt template matching your app category
UI components	AI-generated React/Next.js pages customized to your description
Backend API	FastAPI endpoints with auth, CRUD, and business logic
Database schema	PostgreSQL models based on your features
Styling	Tailwind CSS with professional design patterns
Deployment config	Kubernetes manifests, startup scripts, health checks

This hybrid approach means you get the reliability of tested templates with the customization of AI generation.

3. SmartDoctor Validation

Before deployment, every MVP passes through the SmartDoctor self-healing engine:

Python syntax check — compile() all .py files
JSX/TSX validation — detect unclosed tags, broken imports
Dependency verification — ensure all packages are declared
Config scaffolding — generate missing build configs
Multi-pass repair — detect → fix → verify → repeat

4. Deployment

The final MVP is deployed to Kubernetes:

Gets a live URL: https://mvp-{name}.preview.autoflowly.com
Runs in its own isolated pod
Includes health checks and auto-restart
Frontend served via Nginx, backend via Uvicorn

AI Providers

Autoflowly generates with Claude by default and automatically falls back to OpenAI if the primary provider is unavailable or returns an error — you don't need to configure this yourself.

What Gets Generated

A typical MVP includes 10–25 files:

frontend/
├── src/
│   ├── app/
│   │   ├── layout.tsx          # Root layout with nav
│   │   ├── page.tsx            # Landing page
│   │   ├── login/page.tsx      # Auth pages
│   │   ├── register/page.tsx
│   │   └── dashboard/
│   │       ├── page.tsx        # Main dashboard
│   │       └── settings/page.tsx
│   ├── components/             # Reusable UI components
│   └── lib/                    # Utilities, API client
├── package.json
├── tailwind.config.ts
└── next.config.js

backend/
├── main.py                     # FastAPI app with all routes
├── config.py                   # Environment config
├── requirements.txt            # Python dependencies
└── models/                     # Database models

Customizing After Generation

After your app is generated, you can:

Chat to modify — describe changes in natural language; the AI targets and regenerates only the affected files
Branch it — fork the app into a named branch to try a risky change without touching what's live
Run SmartDoctor — diagnose and fix issues with one click
Re-deploy — push changes live instantly
Hire an AI team — see AI Team & Loops to have agents operate the app going forward

API Access

Generate apps programmatically (requires an authenticated session):

curl -X POST https://api.autoflowly.com/api/conversational-mvp/quick \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <your-token>" \
  -d '{
    "app_name": "Budget Tracker",
    "description": "Personal budget tracker for freelancers with expense categories and monthly reports"
  }'

See the API Reference for full endpoint documentation, or use the MCP integration to build from ChatGPT, Claude, or Gemini directly.