wtf is container-use?

Container Use is an open-source tool that gives AI coding agents isolated environments to work in, so multiple agents can run at the same time without interfering with each other or with your main codebase. It is built by Dagger and is currently in early, experimental development. The core idea is that each agent gets its own fresh container (a self-contained computing environment, similar to a lightweight virtual machine) and its own git branch. When you run multiple agents simultaneously, they stay completely separate: one agent experimenting with a Python refactor will not touch what another agent is doing with the database schema. If an agent makes a mess, you discard that container and branch and nothing else is affected. It works as an MCP server, which is a standard protocol that AI tools like Claude Code and Cursor use to plug in external capabilities. You add Container Use to your agent as an MCP server with one command, and from that point the agent automatically uses containers for its work. You can also drop agent behavior rules into a CLAUDE.md file to guide how the agent interacts with the tool. From your side as a developer, you retain full visibility and control. You can see the complete history of every command an agent ran, not just its summary after the fact. If an agent gets stuck, you can open a terminal directly into its container to see its current state and take over. When an agent finishes, reviewing its work is a standard git checkout of its branch, and merging works the same way you would merge any branch. The tool works with Claude Code, Cursor, Goose, VSCode extensions, and any other agent that supports the MCP protocol. Installation is available through Homebrew on macOS or a shell script for other platforms.