Container

Container, from Aerovato Research, is a command line tool that gives AI coding agents persistent, isolated Linux workspaces on your own machine. Built on Docker, it starts an agent session inside a project specific container, so the agent works in a real Linux environment while destructive actions stay scoped to that project rather than touching the host system. It is aimed at developers who run agents such as Claude Code, OpenCode, Codex and Pi and want them to install packages, run builds and delete files without risk to the rest of the machine. Each project gets its own workspace, and installed tools and runtime state persist between sessions. A guided onboarding step chooses harnesses, tools and defaults, and users can extend the image with their own packages, commands and setup scripts, bring their own distribution and configs, and mount shared configuration across workspaces. Installation is a single script on Windows, macOS or Linux, followed by Docker Desktop, container init and then the container command inside a project directory. No pricing is published. The project's source is on GitHub under the BSD 3 Clause license, and the site invites contributions. The project documentation is candid that container limits what an agent can access but does not make the agent trusted, so review of agent output is still required.