try is a shell tool that lets you run a command and inspect its effects before modifying your live system.
The tool solves the problem of safely testing destructive or system-altering commands by creating an isolated execution environment where changes are not committed to the live system unless explicitly approved. It uses Linux namespaces via unshare and the overlayfs union filesystem to sandbox command execution. Users can review what a command would do, then decide whether to apply those changes to the actual system or discard them.
The tool suits system administrators and developers who need to test commands with potentially significant side effects before running them for real. It works on Linux systems running kernel version 5.11 or higher and has been tested across multiple distributions including Ubuntu, Debian, Fedora, CentOS, Arch, Alpine, Rocky, and SteamOS. The README notes that try is a semisolate rather than a fully isolated sandbox, meaning network calls are allowed and it should only be used with commands you already trust. Installation is flexible: you can grab just the script for minimal setup, clone the repository for full features including documentation and utility support, or download a source distribution from the release page.
The project maintains an active test suite covering multiple Linux distributions and kernel versions. Development includes support utilities designed to improve performance and a manpage for documentation. The tool has been formally evaluated and recognized in academic peer review for its design and implementation approach.
