gpt-engineer
StaleDescription
Platform to experiment with AI Software Engineer — specify software in natural language, watch AI write and execute code, then iterate improvements
Key Features
- Natural language code generation — describe requirements in text, AI writes and executes code automatically
- Iterative improvement — run incremental modification commands on existing codebases with multi-round iteration
- Customizable agent identity — customize AI agent prompts and behavior via preprompts directory
- Vision model input — supports passing images as UX/architecture context for generation
- Multi-backend support — compatible with OpenAI API, Azure OpenAI, and Anthropic models
- Agent benchmarking — built-in bench binary for evaluating agents on APPS/MBPP datasets
Use Cases
💡 Rapidly scaffolding project prototypes from scratch with auto-generated initial code
💡 Running iterative optimization and feature enhancements on existing codebases
💡 Using Vision capability to generate frontend code from design mockups
💡 Switching between multiple LLM backends to test code generation quality
💡 Using built-in benchmarking tools to evaluate custom agent performance
Categories
Quick Start
pip install gpt-engineer
export OPENAI_API_KEY=your-key
mkdir my-project && echo 'Create a REST API with user authentication' > my-project/prompt
gpte my-project