monty

You can safely run Python in a secure sandbox by executing code via uv run with a restricted builtin set, which prevents ordinary filesystem and network access. This isolation is ideal for evaluating LLM-generated snippets and running automated checks.

Yes, you can extend the sandboxed Python environment by loading external safe functions from a user-supplied functions.py file. This allows you to safely expand the sandbox's capabilities without compromising the isolation of the host filesystem or network.

Yes, this tool integrates with uv run to achieve controlled, isolated execution of Python code. This integration provides safety guarantees by restricting builtins and preventing unauthorized network or filesystem access during snippet evaluation.

Using a restricted builtin set limits access to standard library functions that interact with the host environment. To restore specific capabilities while maintaining safe isolation, you must supply external safe functions via a functions.py file.

You need a secure sandbox for running Python code in scenarios requiring safety guarantees, such as evaluating LLM-generated snippets, teaching environments, and automated checks. It prevents unauthorized filesystem and network access during isolated execution.