This project is an agentic AI system that can plan, architect, and generate complete mini web applications from a single natural-language prompt. It uses LangGraph, LangChain, and an LLM to coordinate multiple agents (Planner, Architect, Coder) and produce a fully structured project folder with ready-to-run code.
AppBuilderDemo.mp4
The system follows a multi-agent pipeline:
Converts a user prompt into a structured high-level plan.
Breaks the plan into technical steps and file-level tasks.
Uses tools to automatically create the project files.
AppBuilder/
│
├── main.py # Entry point of the application
├── pyproject.toml # Project dependencies & config
├── uv.lock # Lockfile for reproducible env
│
├── agent/
│ ├── graph.py # LangGraph workflow
│ ├── prompts.py # Prompt templates
│ ├── states.py # State definitions
│ └── tools.py # Custom tools (read/write files, etc.)
│
└── generated_project/ # Auto-generated app output (created at runtime)
- Python 3.10+
uvfor environment & dependency management- An LLM API key (e.g., Groq, OpenAI, or any provider supported by LangChain)
- Clone the repository
git clone https://github.com/faroukbrachemi/AppBuilder.git
cd AppBuilder- Set up the environment
uv init
uv sync- Activate the virtual environment
source .venv/bin/activate- Copy the sample file and add your API key
cp .sample_env .env- Run the project
python main.py- To-Do List App: Create a beautiful to-do list application using HTML, CSS, and JavaScript.
- Calculator App: Create a simple calculator web application using HTML, CSS, and JavaScript.
- Weather App: Generate a small weather app that shows the current weather using HTML, CSS, and JavaScript.
- Quiz App: Create a multiple-choice quiz app using HTML, CSS, and JavaScript.
- Note-Taking App: Build a simple note-taking app using HTML, CSS, and JavaScript.
- Each run generates a fully contained project folder with all required files.
- If you want to generate another project you have to delete the previous one.