antonosika/gpt-engineer

gpt-engineer is a command-line tool that lets you describe a software project in plain English and then uses an AI model to write and execute the code for you. You write your requirements in a simple text file called "prompt," run the tool, and it generates a working codebase. You can also point it at an existing project and describe improvements, and the AI will modify the code accordingly. The project is described as the original proof-of-concept experimentation platform for AI-driven code generation, and is noted as the precursor to Lovable (formerly gptengineer.app), a commercial product built on the same ideas. Under the hood, the tool calls an AI model (OpenAI's models by default, though it also supports Anthropic's Claude and open-source alternatives) with a structured set of instructions called preprompts that define how the AI should behave as an engineer. You can customize these preprompts to shape the AI's identity and make it remember patterns across projects. It also supports passing images alongside text prompts, which is useful for sharing UI mockups or architecture diagrams as additional context. You would use gpt-engineer if you want to quickly prototype a software idea without writing all the code yourself, or if you want to experiment with AI code generation workflows. It also includes a benchmarking tool for testing custom AI agent implementations against standard coding datasets. The tech stack is Python 3.10 through 3.12, installed via pip or Poetry, and it requires an API key for whichever AI provider you choose.