conda-workspaces#

Project-scoped multi-environment workspaces and task runner for conda, with pixi manifest compatibility.

Define environments, features, dependencies, and tasks in a single manifest. conda-workspaces reads conda.toml, pixi.toml, or pyproject.toml and delegates solving and installation to conda — no extra solver, no new package manager, just workspaces and tasks on top of the tools you already use.

Install#

conda

conda install -c conda-forge conda-workspaces

pixi

pixi global install conda-workspaces

Define a workspace#

quickstart demo

Create a conda.toml in your project root:

[workspace]
name = "my-project"
channels = ["conda-forge"]
platforms = ["linux-64", "osx-arm64", "win-64"]

[dependencies]
python = ">=3.10"
numpy = ">=1.24"

[feature.test.dependencies]
pytest = ">=8.0"

[environments]
default = []
test = { features = ["test"] }

Then install and use your environments:

conda workspace install                    # solve + install + generate conda.lock
conda workspace run -e test -- pytest -v   # run a command in an environment
conda workspace shell -e test              # spawn a shell with test env activated
conda workspace install --locked           # reproducible install from conda.lock
conda workspace list                       # list packages in default env
conda workspace envs                       # list defined environments
conda workspace info                       # workspace overview

Define tasks#

task quickstart demo

Add tasks to the same manifest:

[tasks]
test = { cmd = "pytest tests/ -v", depends-on = ["build"] }
build = "python -m build"
lint = "ruff check ."

[tasks.check]
depends-on = ["test", "lint"]

Then run them:

conda task run check        # resolves dependencies, runs build → lint → test
conda task list             # shows all available tasks
conda task run test         # builds first, then tests

Tasks run in your current conda environment by default, or target a workspace environment with -e myenv.

Why conda-workspaces?#

pixi introduced an excellent project model for managing multi-environment workspaces and tasks, but it brings its own solver and installation machinery. conda-workspaces reuses that same manifest format while delegating all solving and installation to conda’s existing infrastructure.

This means:

Workspaces and tasks read from conda.toml, pixi.toml, or pyproject.toml — one manifest, multiple tools
Environments are solved by conda / libmamba and installed as regular conda prefixes
Lock files (conda.lock) capture exact package URLs for reproducible installs without re-solving
Task dependencies, caching, Jinja2 templates, and platform overrides all work out of the box
Ships as a conda plugin (conda workspace, conda task) and standalone conda workspace / conda task CLIs (also available as cw / ct)