从想法到交付,只需一个脚本。体验真正的“解放双手”。
Liberate your hands. Experience true hands-free software development. Auto_Claude_CLI is a headless software factory powered by a single script. It orchestrates a virtual team of AI agents (PM, Architect, Developer, QA) to transform raw ideas into fully functional, refactored, and tested codebases while you sleep.
Run full autonomous cycles for < $0.10 / day.
Don't overpay for coding. I highly recommend using the GLM-Coding model with Claude CLI for the best cost-to-performance ratio.
👉 Try GLM-Coding (High Performance, Low Cost)
Signing up via this link reduces your API costs and directly supports my open-source work!
Get up and running Now. No complex environment setup required.
# Windows
curl https://raw.githubusercontent.com/vincentzzh424/auto_claude_cli/main/run.py -o run.py; python run.py "write a test print hello world"
# Mac&Linux
curl -sL https://raw.githubusercontent.com/vincentzzh424/auto_claude_cli/main/run.py | python - "write a test print hello world"Install the underlying driver (Node.js required).
npm install -g @anthropic-ai/claude-codeOption A (Recommended): Get GLM-Coding Key (~$0.1/day)
Option B: Use Official Anthropic Key ($20/month)
Dependency-free , minimal python requirements.
git clone https://github.com/vincentzzh424/auto_claude_cli.git
cd auto_claude_cli
# (Optional) Create a workspace to keep things clean
mkdir demo && cd demo # Run the script from the parent directory (or add to path) with your idea.
python ../run.py "Create a Personal Website with a dark mode toggle"
💡 Usage Examplespython ../run.py "Create a CLI tool to resize and watermark images in a folder"💡Complex System Tip: Use ChatGPT/Gemini to refine your prompt into a detailed specification before pasting it here.
python ../run.py "Build a Bloomberg-like terminal for crypto trading. Features: WebSocket data feed for BTC/ETH, real-time K-line visualization using mplfinance, and MACD/RSI indicator calculation. No mock data."Please Read Before Use
To achieve true "hands-free" autonomy, this tool executes the Claude CLI with the --dangerously-skip-permissions flag enabled.
Risk: The agent has full filesystem access (read/write/delete).
Recommendation: Always run this tool inside a Sandbox, Docker container, or a dedicated empty directory to prevent accidental data loss.
We believe in the power of autonomous agents. Pull Requests to improve the driver logic or prompt engineering are welcome.
📄 License MIT © [vincentzzh424]