code-on-incus

Why we included this project

AI coding agents like Claude Code, Codex, and opencode need room to install packages and run commands, which is a lot of trust to hand over on a machine that holds your credentials. code-on-incus (coi) sidesteps that by giving each agent its own full Linux container through Incus, with root, systemd, and Docker inside, so it behaves like a real server while your host stays out of reach. Your project mounts with correct file ownership, and SSH keys, tokens, and environment variables never enter the container unless you explicitly allow them. The active defense layer is the part that stands out: kernel-level monitoring watches for reverse shells, data exfiltration, and credential scanning, and pauses or kills the container on its own when it detects something dangerous. For anyone who runs coding agents regularly and wants real isolation plus visibility into their actions, this is a self-hosted way to get both without babysitting.