gpt-engineer
Command-line tool that asks an AI model to write a project from a plain-language prompt.
- Stars
- 55k
- Forks
- 7.3k
- Licence
- MIT
- Language
- Python
- Last update
- 1.5 years ago
- Type
- CLI
This repository is archived and no longer maintained.
What it is
gpt-engineer is a command-line tool where you describe a piece of software in a text file called prompt, and it uses an AI model (such as OpenAI's models) to write and run the code. The README calls it an experimentation platform for code generation, and says it is the precursor to the commercial gptengineer.app. The project is archived on GitHub, so it is no longer actively developed.
- Generate a small project from a written description by running gpte on an empty folder
- Ask the tool to improve an existing project folder with the -i option
- Provide UX or architecture diagrams as image input with a vision-capable model
- Benchmark your own agent against the APPS or MBPP datasets using the bench binary
Great for
- Developers who want to experiment with AI-written code from a simple prompt
- Students exploring how code generation agents work
- Researchers or builders who want a hackable base for their own coding agent
- People who already have an OpenAI or Anthropic API key
Not ideal for
- People who want a maintained, production-ready tool: the repository is archived and the README points to aider for a well-maintained CLI
- Non-technical users who need a visual interface: the README describes the CLI only and points to gptengineer.app for a managed UI
- Anyone unwilling to pay for model API usage, since the README requires an API key or a custom model setup
How to get it
- 1
Install with a package manager
You'll need: Python 3.10 to 3.12 · An OpenAI API key or another supported model setup
Install the stable release
python -m pip install gpt-engineer - 2
Run it
- Set your API key with export OPENAI_API_KEY=[your api key] or via a .env file
- Create an empty project folder and add a file named prompt with your instructions
- Run gpte <project_dir> with the relative path to your folder
- Review the generated files in the project folder
- 3
Get the most out of it
- Edit the preprompts folder or use --use-custom-preprompts to change the agent's behavior
- Use the -i flag to ask the AI to improve existing code
- Add image inputs with --image_directory and a vision-capable model
- Use the bench binary and the template repo to benchmark custom agents
- Use the documented open model setup to run with local or alternative models
Downloads
Strengths
- Simple workflow: write a prompt file and run one command
- Open source under the MIT license with a large community (about 55k stars)
- Customizable through preprompts and support for alternative models
Trade-offs
- The repository is archived, so updates and fixes are unlikely
- Results depend on the underlying AI model and need manual review of generated code
- Requires an API key or extra setup for local or open models
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| gpt-engineer (this) | 55k | CLI | intermediate | 1.5 years ago | Developers who want to experiment with AI-written code from a simple prompt |
| OpenHands | 92.5k | App | intermediate | 4 days ago | Developers who want coding agents running on a server while their laptop is off |
| oh-my-pi | 37.5k | CLI | intermediate | 4 days ago | Developers who want an AI agent working inside a terminal on their own codebase |
| plandex | 15.9k | CLI | intermediate | 1.1 years ago | Developers working in large codebases who want an AI agent to handle multi-file changes |
| shell_gpt | 12.3k | CLI | beginner | 4 months ago | Developers who often forget shell or command-line syntax |
| SuperAGI | 17.7k | AI model / agent | advanced | 1.8 years ago | Python developers experimenting with autonomous AI agents |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.