micromamba

Create environment

$ micromamba create -n [myenv] python=[3.11]

$ micromamba activate [myenv]

$ micromamba install -n [myenv] [numpy] [pandas]

$ micromamba install -c conda-forge [package]

$ micromamba env list

$ micromamba list -n [myenv]

$ micromamba env remove -n [myenv]

$ micromamba update --all

micromamba command [-n env] [-c channel] [options] [packages]

micromamba is a tiny, statically linked reimplementation of the conda package manager. It provides the same functionality as mamba/conda but as a single, fast, standalone binary.
Unlike conda/mamba, micromamba requires no base environment or Python installation. It's a pure C++ implementation that can bootstrap an entire conda environment from scratch.
Package resolution uses the same libsolv library as mamba, providing fast dependency solving. Channels like conda-forge provide access to thousands of packages.
Environment management mirrors conda: create isolated environments, install packages, and switch between them. Environment files (YAML) enable reproducible environment creation.
The tool is particularly useful for CI/CD pipelines and containers where a minimal footprint is desired. No initialization or base install needed - just download and run.
Shell integration works with bash, zsh, fish, and PowerShell. The activate command modifies PATH and environment variables appropriately.

create

Create new environment.

Activate environment.

Deactivate current environment.

Install packages.

Remove packages.

Update packages.

List installed packages.

List environments.

Remove environment.

Search for packages.

Clean package cache.

Environment name.

Package channel.

Environment file.

Don't ask for confirmation.

Don't install dependencies.

Quiet output.

Some conda features not implemented. No graphical interface. Some older packages may have compatibility issues. Shell hooks must be initialized. Different default channel than conda (uses conda-forge).

micromamba was developed by the mamba team at QuantStack, released around 2020. It was designed for scenarios where conda's size and startup time were problematic. The project shares its solver with mamba, bringing conda-like capabilities to constrained environments.

conda(1), mamba(1), pip(1), pixi(1)