Introduction

MANYOYO lets Claude Code / Codex / Gemini / OpenCode run in YOLO mode inside a container: the AI can go wild without being able to hurt your computer.

Why you need it

AI agent CLIs read and write your code, run shell commands and install dependencies. The no-confirmation (YOLO / SOLO) modes are the most productive, but running them directly on your machine leaves the risk boundary unclear. MANYOYO puts them in a container sandbox: the agent only sees the work directory you give it, and a broken container can be deleted and recreated at any time.

What it does

• Multiple agents: claude, codex, gemini and opencode behind one entry point, switch any time.
• Container isolation: built on Docker / Podman to reduce what the host exposes.
• Web UI: chat, browse files and open a terminal in the browser, phones included.
• One configuration: run profiles, environment variables, mounts and image options live in ~/.manyoyo/manyoyo.json.
• One command to install, upgrade and uninstall.

Aspect	Bare agent CLI	MANYOYO
Host exposure	High	Lower
Run boundary	Scattered	Centralized in container and config
Reproducibility	Weak	Strong (image + config)
High-risk mode notes	Left to each tool	Explicit YOLO / SOLO / sock warnings

Security boundary

MANYOYO reduces risk but is not "absolutely safe"; see Security Notes.

Next Steps

• Install
• First Run