uv: one tool to replace pip, venv, pyenv and pipx for Python, many times faster

Starting a Python project used to take a whole toolbox: pyenv to install Python versions, virtualenv for environments, pip to install packages, pip-tools or Poetry to lock versions, pipx for command-line tools. Each has its own conventions, and installing dependencies for a large project can take minutes. uv from Astral, written in Rust, replaces that entire set with one command, and according to its README is 10 to 100 times faster than pip.

GitHub: https://github.com/astral-sh/uv · about 91k stars · Rust · Apache-2.0/MIT

Installation

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
# Windows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
# Or
pipx install uv     # pip install uv
brew install uv

Managing a project

uv init my-api          # create a project: pyproject.toml, .python-version, main.py
cd my-api
uv add fastapi uvicorn  # add dependencies, update pyproject.toml and uv.lock
uv add --dev pytest     # development dependency
uv run main.py          # run inside the project environment (creates .venv if needed)
uv run pytest
uv sync                 # install exactly what uv.lock says, e.g. after cloning
uv lock --upgrade       # upgrade versions in the lock file
uv remove uvicorn

No need to "activate" a virtualenv: uv run always runs in the project's environment and syncs dependencies before running.

What does uv replace?

The uv pip interface is compatible with familiar pip commands, so you can migrate older projects gradually without restructuring them right away.

Single-file scripts with dependencies

uv supports declaring dependencies at the top of a script file (inline metadata):

uv init --script fetch.py
uv add --script fetch.py requests rich
uv run fetch.py        # uv creates a temporary environment with requests and rich

Great for utility scripts shared across a team: send one file and it runs, no setup instructions needed.

Caveats

• Commit both pyproject.toml and uv.lock; in CI use uv sync --locked to fail if the lock file is out of date.
• Python versions installed by uv live in uv's own directory and don't touch the system Python.
• For Poetry or pip-tools projects that already work, migrate step by step and double-check dependency versions after locking.

How to apply it

• Dev: Use uv init for new projects, uvx for one-off tools, and uv run instead of activating venvs.
• Tech Lead: Standardise on uv for the team and CI: one tool, one lock file, and much faster installs mean shorter pipelines.
• PM: Python environment setup for new developers drops from "half a day" to a few commands; faster CI also saves money.