
DeepSeek Coder
Let the Code Write Itself

CLI platform to experiment with code generation using natural language.

GPT Engineer is an open-source CLI platform designed for experimenting with code generation. It enables users to specify software requirements in natural language, which the AI then uses to write and execute code. It supports different models including OpenAI, Azure OpenAI, and open-source models like WizardCoder. The tool can improve existing code by accepting instructions via a prompt file and allows users to benchmark custom agent implementations against datasets. Key features include overriding preprompts for agent identity and accepting image inputs for vision-capable models. It's a precursor to gptengineer.app, a commercial project with a UI, and Aider, a well-maintained CLI tool.
GPT Engineer is an open-source CLI platform designed for experimenting with code generation.
Explore all tools that specialize in natural language input. This domain focus ensures GPT Engineer delivers optimized results for this specific requirement.
Override the default preprompts folder to specify the 'identity' of the AI agent. This allows for persistent memory and behavior across projects.
Accept image inputs alongside text prompts to provide additional context, such as UX or architecture diagrams, for vision-capable models.
Supports OpenAI, Azure OpenAI, Anthropic, and open-source models like WizardCoder, offering flexibility in model selection.
Includes a 'bench' binary for benchmarking custom agent implementations against public datasets like APPS and MBPP.
Provides Docker instructions for easy deployment and consistent execution across different environments.
Install gpt-engineer: `python -m pip install gpt-engineer` (for stable release) or clone the repo for development.
Install poetry: `poetry install` within the cloned directory.
Activate the virtual environment: `poetry shell`.
Set up your API key by exporting it as an environment variable: `export OPENAI_API_KEY=[your api key]` or create a `.env` file.
Create a project directory and a 'prompt' file inside it with instructions.
Run gpt-engineer: `gpte <project_dir>` to generate code or `gpte <project_dir> -i` to improve existing code.
All Set
Ready to go
Verified feedback from other users.
"Generally positive reviews highlighting its usefulness for rapid prototyping and experimentation, but also noting the need for careful review and refinement of the generated code."
Post questions, share tips, and help other users.

Let the Code Write Itself

A pre-trained model for programming and natural languages.

AI-powered coding assistant that builds apps and websites from natural language prompts.

Build real-world software with AI using an open-source, terminal-based coding agent.

A 15B parameter model trained on 600+ programming languages, designed for code generation and understanding.

The modern coding superpower: ultra-fast AI autocomplete, chat, and repository-wide context.