#Installation

Install molt-release once and you get a molt command on your PATH; then run molt init to set up a project.

#Requirements

• Python 3.11 or newer. molt reads TOML with the standard-library tomllib parser, which landed in 3.11. The floor stops there so molt runs in the CI you already have.
• A workspace on any supported ecosystem. uv is first-class; Poetry, Hatch, PDM, and setuptools are covered through a backend seam. See Ecosystems.
• Windows, macOS, or Linux. molt is tested on Windows from the first commit and keeps its output ASCII-clean, so it behaves the same in every runner.

#Install as a tool

The recommended path is uv tool install, which puts molt in its own isolated environment and exposes the molt command globally:

uv tool install molt-release

Prefer not to install anything? Run it straight from the index with uvx:

uvx molt-release --help

Either way, the command you type afterward is always molt:

molt --help

Other installers work too, since molt-release is an ordinary PyPI distribution:

pipx install molt-release
# or, into the current environment
pip install molt-release

- uses: giancarlosisasi/molt-action@v1
  with:
    molt-version: "0.1.3"
    publish: molt publish

#molt-release the package, molt the command

You install molt-release but you run molt. The distribution is named molt-release because molt on PyPI belongs to an unrelated project. A distribution name and the console script it installs are independent, so molt-release still gives you a molt binary. You never type molt-release again after installing. See the Overview.

#Verify the install

molt --version

This prints the bare version string (for example, 0.1.1) and touches none of your project -- it is safe to run anywhere.

#Scaffold a project

From the root of the repository you want to manage, run:

molt init

molt init creates a .changeset/ directory -- the home for the changeset files you will write -- and adds a [tool.molt] configuration block. It is safe to re-run: if molt is already set up, it reports that and changes nothing.

.changeset/
  README.md          # a short guide, committed alongside your changesets
pyproject.toml       # now contains a [tool.molt] section

See molt init for the flags and prompts, and The config file for where configuration lives and every option it accepts.

#Where to go next

• One package? Follow the single-package quickstart -- the majority Python case, and first-class in molt.
• A workspace? Follow the monorepo quickstart to see dependency propagation in action.
• Want the mental model first? Read The release workflow.
• Releasing from GitHub? Set up the GitHub Action and molt runs molt version and molt publish for you, on every push to your base branch.
• Coming from JavaScript changesets? Molt and changesets maps the commands, configuration keys, and changeset fields one to one, and Migrating from changesets converts an existing .changeset/ directory.