A Smoother, Faster Way to Run AI Agents with Container Use

Container Use is a tool that gives AI agents real-world superpowers: the ability to run shell commands in parallel, persist environments, and act like real developers. It’s how you bring LLMs closer to doing useful, repeatable work.

This release focuses on making that experience smoother, faster, and easier to trust.

What's better today than yesterday?

You can reuse environments without starting over

Want your agent to continue where another left off? You can now prompt it using an existing environment ID:
using the existing env clever-tiger, do ...
No need to re-bootstrap or re-run past commands. It just works.

You can inspect what the agent did clearly

Ever wonder what your agent actually produced?
In addition to cu log <env>, with the new cu checkout <env> command, you can instantly inspect its most recent output, no digging required.

You spend less time typing and guessing

Shell autocompletion is now working and auto-installed via Homebrew.
That means faster workflows, fewer typos, and less time looking up command syntax.

You resume work right where you left off

Open a terminal session, and everything is back: your environment, your commands, your context.
It's like the agent never left.

You don’t have to worry about edge cases

Empty files are now persisted correctly. Environment IDs are easier to remember and reference. Behind the scenes, everything is simpler and lighter, but you won’t need to think about it.

Real improvements, made possible by real contributors

This release was made better thanks to contributions from our community:

• @zhichli
• @InTheCloudDan
• @michaelneale

Thank you for your PRs, feedback, and support. You’re helping make Container Use a better tool for everyone building with agents.

Ready to try it?
Get started here → https://github.com/dagger/container-use
Ask questions in Discord or create an issue