This Sandbox Tool Makes Claude Code Unstoppable (Code On Incus)

Description

Check out how to run autonomous AI agents safely on your Mac by creating an isolated "Inception-style" sandbox using Colima, Incus, and Claude Code. This tutorial walks you through solving common security risks and networking hurdles to build a collaborative multi-agent system that can develop apps and analyze files without ever touching your host machine. Whether you're building software or inspecting suspicious files, this setup provides the ultimate balance of AI power and impenetrable container security. 🔗 Relevant Links Code On Incus: https://github.com/mensfeld/code-on-incus Incus: https://linuxcontainers.org/incus/ Colima: https://github.com/abiosoft/colima ❤️ More about us Radically better observability stack: https://betterstack.com/ Written tutorials: https://betterstack.com/community/ Example projects: https://github.com/BetterStackHQ 📱 Socials Twitter: https://twitter.com/betterstackhq Instagram: https://www.instagram.com/betterstackhq/ TikTok: https://www.tiktok.com/@betterstack LinkedIn: https://www.linkedin.com/company/betterstack 📌 Chapters: 00:00 Introduction: The AI Security Problem 01:01 What is Incus? 01:21 How Code on Incus Works 01:46 Persistent State & UID Mapping 02:18 Setup: Running Incus on Mac 03:01 Installing & Configuring Incus 03:45 Network & Firewall Fixes 04:37 Building the 'coi' Tool 05:13 Demo 1: Multi-Agent Collaboration 06:45 Demo 2: Secure Malware Analysis 08:43 Final Thoughts & Conclusion