Agent-Yes! for Claude/Codex/Gemini/Cursor/Copilot/Qwen/Auggie
A wrapper tool that automates interactions with various AI CLI tools by automatically handling common prompts and responses. Originally designed for Claude CLI, now supports multiple AI coding assistants. Rewritten in Rust for improved performance and reliability.
⚠️ Important Security Warning: Only run this on trusted repositories. This tool automatically responds to prompts and can execute commands without user confirmation. Be aware of potential prompt injection attacks where malicious code or instructions could be embedded in files or user inputs to manipulate the automated responses.
Install
One-liner (installs Bun if needed, then the ay / cy / claude-yes … CLIs):
# macOS / Linux
curl -fsSL https://agent-yes.com/setup.sh | sh# Windows (works in both PowerShell and cmd)
powershell -c "irm https://agent-yes.com/setup.ps1 | iex"Or with a package manager you already have:
bun add -g agent-yes # or: npm install -g agent-yesThen: ay claude (run an agent with auto-yes) · ay serve --share (web console + shareable link) · live console at https://agent-yes.com
For the local web console, install Portless once with npm install -g portless, then run ay serve. It assigns a free internal port and serves the console at https://agent-yes.localhost/. ay serve --port N remains available for a fixed-port API listener.
Features
- Multi-CLI Support: Works with Claude, Gemini, Codex, Copilot, and Cursor CLI tools
- Auto-Response: Automatically responds to common prompts like "Yes, proceed" and "Yes"
- Continuous Operation: Keeps the AI assistant running until your task is done, waiting for your next prompt
- Interactive Control: You can still queue more prompts or cancel executing tasks with ESC or Ctrl+C
- Crash Recovery: Automatically restarts crashed processes (where supported)
- Idle Detection: Optional auto-exit when the AI becomes idle
- Named Pipe Input (Linux): On Linux systems, automatically creates a FIFO (named pipe) at /tmp/agent-yes-YYYYMMDDHHMMSSXXX.stdin for additional input streams
- Isolated Processes: Each ay <cli is an independent wrapper — no central daemon owns the agents. They coordinate through files (pids.jsonl index, per-pid FIFO for stdin, per-cwd .raw.log for output), so one process crashing (even ay serve) never takes down the others. See docs/architecture.md.
Agent Clis
Install the AI CLI tool(s) you want to use:
Claude
npm install -g @anthropic-ai/claude-code