genStack — AI-Powered Full-Stack Project Generator

genStack is an open-source web application that generates complete, production-ready full-stack projects from a single text prompt. Powered by Anthropic Claude or Google Gemini (your choice), it uses MCP (Model Context Protocol) tool integration to intelligently create, modify, and deploy multi-language codebases.

---

💰 Donate — Support this project

If you find this project useful, consider supporting it!

Scan the QR code below to donate via UPI:

UPI ID: 8778990202a@kotak

---

Quick Start (TL;DR)

git clone https://github.com/nanunh/genstack.git
cd genstack
python3 -m venv venv && source venv/bin/activate   # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env                               # fill in your API key + DB details
python3 generate_keys.py                           # generates SSL certificates
python3 server.py                                  # starts the server

Then open https://localhost:8000 in your browser.

On first visit you will see a browser security warning — click Advanced → Proceed to localhost. This is expected for self-signed certificates.

---

LLM Provider Setup

genStack supports Anthropic Claude and Google Gemini. You only need one API key. Set the values in your .env file.

---

Option A — Anthropic Claude

Get your API key at: https://console.anthropic.com/settings/keys

# .env
ANTHROPIC_API_KEY=sk-ant-api03-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

# Optional: override the default model (default: claude-sonnet-4-5-20250929)
# ANTHROPIC_MODEL=claude-sonnet-4-5-20250929

genStack auto-detects Anthropic when ANTHROPIC_API_KEY is present. No LLM_PROVIDER line needed.

Recommended Anthropic models:

Model	ID	Best for
Claude Sonnet 4.5 (default)	claude-sonnet-4-5-20250929	Best balance of quality and speed
Claude Sonnet 4.6	claude-sonnet-4-6	Latest, highest quality
Claude Haiku 4.5	claude-haiku-4-5-20251001	Faster and cheaper

To switch model, add to .env:

ANTHROPIC_MODEL=claude-sonnet-4-6

---

Option B — Google Gemini

Get your API key at: https://aistudio.google.com/app/apikey

# .env
GEMINI_API_KEY=AIzaSy-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

# Optional: override the default model (default: gemini-2.0-flash)
GEMINI_MODEL=gemini-2.5-flash

genStack auto-detects Gemini when GEMINI_API_KEY is present (and no Anthropic key is set). No LLM_PROVIDER line needed.

Recommended Gemini models:

Model	ID	Best for
Gemini 2.5 Flash (recommended)	gemini-2.5-flash	Best quality, 65K output tokens
Gemini 2.0 Flash (default)	gemini-2.0-flash	Fast, 8K output tokens
Gemini 2.5 Pro	gemini-2.5-pro	Highest quality

To switch model, add to .env:

GEMINI_MODEL=gemini-2.5-flash

Note: Use gemini-2.5-flash or gemini-2.5-pro for generating larger projects. The older gemini-2.0-flash has an 8 192-token output limit which can truncate complex projects.

---

Option C — Force a specific provider

If you have both keys set, Anthropic takes priority by default. To override, set LLM_PROVIDER explicitly:

# Use Anthropic even if GEMINI_API_KEY is also present
LLM_PROVIDER=anthropic

# Use Gemini even if ANTHROPIC_API_KEY is also present
LLM_PROVIDER=gemini

---

Complete .env examples

Using Anthropic:

ANTHROPIC_API_KEY=sk-ant-api03-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
ANTHROPIC_MODEL=claude-sonnet-4-6

DB_HOST=localhost
DB_PORT=3306
DB_NAME=stackgpt_db
DB_USER=stackgpt_user
DB_PASSWORD=your_db_password

Using Gemini:

GEMINI_API_KEY=AIzaSy-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
GEMINI_MODEL=gemini-2.5-flash
LLM_PROVIDER=gemini

DB_HOST=localhost
DB_PORT=3306
DB_NAME=stackgpt_db
DB_USER=stackgpt_user
DB_PASSWORD=your_db_password

Using both keys (Anthropic active, Gemini as fallback option):

ANTHROPIC_API_KEY=sk-ant-api03-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
GEMINI_API_KEY=AIzaSy-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
LLM_PROVIDER=anthropic   # switch to "gemini" any time without removing keys

---

Features

• Prompt-to-Project — Describe your app in plain English and get a complete project with all files generated
• Dual LLM Support — Works with Anthropic Claude and Google Gemini; switch providers by changing one env variable
• Multi-Language AST Parsing — Understands and modifies code structure across Python, JavaScript, TypeScript, and more via Tree-sitter
• AI Code Assistant — Chat with the AI to add features, fix bugs, create/delete/modify files in your generated project
• Token Usage Dashboard — Tracks and estimates API cost per project and operation (provider-aware pricing)
• User Authentication — JWT-based signup/login with MySQL backend
• SSH Deployment — Deploy generated projects directly to a remote Linux server via SSH/SCP with PM2 support
• Project History — Browse and reload previously generated projects
• File Upload & Analysis — Upload existing code files for AI analysis and enhancement

---

Tech Stack

Layer	Technology
Backend	Python 3.10+, FastAPI, Uvicorn
AI (Anthropic)	Claude Sonnet 4.5 / 4.6 via anthropic SDK
AI (Gemini)	Gemini 2.0 / 2.5 Flash via google-generativeai SDK
AST Parsing	Tree-sitter
Auth	JWT (python-jose), bcrypt (passlib)
Database	MySQL 8.0+
Deployment	Paramiko (SSH), SCP, PM2
Frontend	Vanilla HTML / CSS / JavaScript

---

Prerequisites

1. Python

Install Python 3.10 or higher:

• Windows: https://www.python.org/downloads/windows/
• macOS: https://www.python.org/downloads/macos/
• Linux (Ubuntu/Debian): https://docs.python-guide.org/starting/install3/linux/

Verify installation:

python --version

2. MySQL

Install MySQL 8.0 or higher:

• Windows: https://dev.mysql.com/downloads/installer/
• macOS: https://dev.mysql.com/downloads/mysql/
• Linux (Ubuntu/Debian): https://dev.mysql.com/doc/refman/8.0/en/linux-installation.html

Verify installation:

mysql --version

3. An LLM API Key

You need one of the following:

Provider	Where to get it	Cost
Anthropic	https://console.anthropic.com/settings/keys	Pay-per-token
Google Gemini	https://aistudio.google.com/app/apikey	Free tier + pay-per-token

---

Installation

1. Clone the repository

git clone https://github.com/nanunh/genstack.git
cd genstack

2. Create a virtual environment

Windows:

python -m venv venv
venv\Scripts\activate

macOS / Linux:

python -m venv venv
source venv/bin/activate

3. Install Python dependencies

pip install -r requirements.txt

4. Generate SSL Certificates

genStack runs over HTTPS and requires SSL certificate files in the keys/ folder. Run the included key generator to create self-signed certificates for local development:

python3 generate_keys.py

What this does:

• Generates an RSA 4096-bit private key → keys/privkey.pem
• Generates a self-signed X.509 certificate → keys/fullchain.pem
• Valid for 825 days (~2 years)
• Covers localhost, 127.0.0.1, and ::1

Expected output:

✅ SSL keys generated successfully!
  Private key : keys/privkey.pem
  Certificate : keys/fullchain.pem
  Valid until : 2028-06-01

Browser warning: Self-signed certificates will trigger a security warning in your browser. Click Advanced → Proceed to localhost (unsafe) to continue. This is safe for local development.

Production use: Replace keys/privkey.pem and keys/fullchain.pem with real certificates from Let's Encrypt using certbot:

sudo certbot certonly --standalone -d yourdomain.com
sudo cp /etc/letsencrypt/live/yourdomain.com/privkey.pem keys/privkey.pem
sudo cp /etc/letsencrypt/live/yourdomain.com/fullchain.pem keys/fullchain.pem

5. Configure environment variables

Copy the example env file and fill in your values:

cp .env.example .env

Open .env and set your LLM provider key and database credentials. See the LLM Provider Setup section above for full examples.

Minimum required fields:

# Pick ONE provider:
ANTHROPIC_API_KEY=sk-ant-...    # Anthropic Claude
# or
GEMINI_API_KEY=AIzaSy-...       # Google Gemini

# Database
DB_HOST=localhost
DB_PORT=3306
DB_NAME=stackgpt_db
DB_USER=your_mysql_user
DB_PASSWORD=your_mysql_password

---

Database Setup

Step 1 — Create a MySQL user (optional but recommended)

Log in to MySQL as root:

mysql -u root -p

Run the following SQL:

CREATE USER 'stackgpt_user'@'localhost' IDENTIFIED BY 'your_password';
GRANT ALL PRIVILEGES ON stackgpt_db.* TO 'stackgpt_user'@'localhost';
FLUSH PRIVILEGES;
EXIT;

Update DB_USER and DB_PASSWORD in your .env to match.

Step 2 — Auto-initialize the database

The database and all required tables are created automatically the first time you run the server. No manual SQL import is needed.

Tables created on first run:

• users — stores registered user accounts
• user_sessions — stores active sessions
• code_assistant_history — stores per-project chat history

If you need to recreate the tables manually, you can trigger it by running:

python database.py

---

Running the Server

Make sure you have completed all setup steps above (including generate_keys.py), then run:

python3 server.py

Then open your browser at:

https://localhost:8000

First-time browser warning: Click Advanced → Proceed to localhost to bypass the self-signed certificate warning. This only appears once per browser session.

You will be greeted by the login/signup page. Create an account and start generating projects.

The startup log confirms which LLM provider is active:

LLM provider : gemini (model: gemini-2.5-flash)

or

LLM provider : anthropic (model: claude-sonnet-4-5-20250929)

---

Project Structure

genstack/
├── server.py                  # FastAPI application entry point
├── generate_keys.py           # SSL key pair generator (run before server.py)
├── auth.py                    # JWT authentication helpers
├── database.py                # MySQL connection & table initialization
├── store.py                   # LLM client initialization & provider detection
├── models.py                  # Pydantic data models
├── multiLanguageASTParser.py  # Tree-sitter multi-language AST parser
├── ast_cache_manager.py       # AST grammar cache manager
├── enhanced_ast_modifier.py   # AI-driven AST code modifier
├── token_usage_manager.py     # Token usage tracking & cost estimation
├── requirements.txt           # Python dependencies
├── .env.example               # Environment variable template
├── services/
│   ├── llm_provider.py        # Unified Anthropic + Gemini streaming abstraction
│   ├── project_generator.py   # MCP-based project generation logic
│   ├── code_assistant.py      # AI code modification & chat logic
│   └── mcp_tools.py           # MCP tool definitions and execution
├── routes/                    # FastAPI route handlers
├── keys/                      # SSL certificates (git-ignored)
│   ├── privkey.pem
│   └── fullchain.pem
├── static/                    # Frontend (HTML, CSS, JS)
│   ├── index.html
│   ├── login.html
│   ├── signup.html
│   ├── app.js
│   ├── auth.js
│   ├── styles.css
│   └── auth-styles.css
├── generated_projects/        # Runtime: AI-generated project output (git-ignored)
├── token_usage/               # Runtime: token usage data (git-ignored)
└── ast_cache/                 # Runtime: compiled AST grammars (git-ignored)

---

Environment Variables Reference

Variable	Required	Default	Description
ANTHROPIC_API_KEY	One of these two	—	Anthropic API key
GEMINI_API_KEY	One of these two	—	Google Gemini API key
LLM_PROVIDER	No	auto-detected	Force provider: anthropic or gemini
ANTHROPIC_MODEL	No	claude-sonnet-4-5-20250929	Override Anthropic model
GEMINI_MODEL	No	gemini-2.0-flash	Override Gemini model
SERVER_URL	No	https://localhost:8000	Server base URL
OUTPUT_DIR	No	generated_projects	Generated projects directory
LOG_LEVEL	No	INFO	Logging verbosity
DB_HOST	Yes	localhost	MySQL host
DB_PORT	No	3306	MySQL port
DB_NAME	Yes	—	MySQL database name
DB_USER	Yes	—	MySQL username
DB_PASSWORD	Yes	—	MySQL password

---

License

MIT License — see LICENSE for details.

Copyright (c) 2026 River Transformation Hub