Few concepts are extensively used in Mamba and in this documentation as well. You should start by getting familiar with those as a starting point.


In Unix-like platforms, installing a software consists in placing files in subdirectories of an “installation prefix”:

  • no file is placed outside of the installation prefix

  • dependencies must be installed in the same prefix (or standard system prefixes with lower precedence)


Examples on Unix: the root of the filesystem, the /usr/ and /usr/local/ directories.

A prefix is a fully self-contained and portable installation.
To disambiguate with root prefix, prefix is often called target prefix. Without explicit target or root reference, you can assume it points a target prefix.

An environment is just another way to call a target prefix.

Mamba’s environments are similar to virtual environments known from Python’s virtualenv and similar software, but more powerful since Mamba also manages native dependencies and generalizes the virtual environment concept to many programming languages.

Root prefix#

When downloading for the first time the index of packages for resolution of the environment, or the packages themselves, a cache is generated to speed-up next operations:

  • the index has a configurable time-to-live (TTL) during which it will be considered as valid

  • the packages are preferentially hard-linked to the cache location

This cache is shared by all environments or target prefixes based on the same root prefix. Basically, that cache directory is a subdirectory located at $root_prefix/pkgs/.

The root prefix also provide a convenient structure to store environments $root_prefix/envs/, even if you are free to create an environment elsewhere.

Base environment#

The base environment is the environment located at the root prefix.

This is a legacy environment from conda implementation still heavily used.
The base environment contains the conda and mamba installation alongside a Python installation (since mamba and conda require Python to run)
mamba and conda being themselves Python packages, they are installed in base environment, making the CLIs available in all activated environment based on this base environement.


You can’t create base environment because it’s already part of the root prefix structure, directly install in base instead.



The activation of an environment makes all its contents available to your shell. It mainly adds target prefix subdirectories to your $PATH environment variable.


activation implementation is platform dependent.

When activating an environment from another, you can choose to stack or not upon the currently activated env.
Stacking will result in a new intermediate prefix: system prefix < base < env1 < env2.


The deactivation is the opposite operation of activation, removing from your shell what makes the environment content accessible.