Using uv for Python
When running Python code or managing Python dependencies, always use uv instead of pip or python directly.
uv should be installed - if it is not, ask the user to brew install uv or similar depending on their OS. Offer to do it on their behalf.
If you are running in sandbox mode, tell the user that you want to add the uv cache directory to the list of allowed paths, by adding the snippet below to the PROJECTDIR/.claude/settings.local.json file. Once this is done, uv will be able to read and write its cache even in sandbox mode.
{
"permissions": {
"allow": [
"Read(~/.cache/uv)",
"Edit(~/.cache/uv)",
// other allowed paths...
],
}
}
Preferred python version
The preferred default python version to use and set up for projects is 3.12.
Write code using type hints.
For projects or very complex standalone scripts, set up ruff (linter/formatter) and ty (type checker) as dev dependencies and run both when you make changes to python source code.
Python projects vs standalone scripts
For python projects with dependencies, always use uv to manage dependencies (uv init, uv add, uv sync, etc) and run scripts (uv run).
For standalone scripts, use comment headers to specify dependencies like so:
#!/usr/bin/env -S uv run --script # /// script # requires-python = ">=3.12" # dependencies = [ # "requests", # ] # /// # rest of your python script comes below
Running Python standalone scripts
Instead of:
python script.py
Use:
uv run --script script.py
Running third party tools/scripts
For one-off tool execution:
uv tool run <tool-name> # or shorthand uvx <tool-name>
Extended documentation
For complex interactions with uv or to debug uv issues, refer to the online documentation at https://docs.astral.sh/uv/llms.txt and the pages linked from there. This is not necessary to just run scripts, add dependencies etc.