conda-lock
conda.github.io
- Free plan
- Windows
- macOS
- Linux
What conda-lock is
conda-lock is a free library for producing reproducible lockfiles for conda environments. It resolves dependencies separately for each target platform, allowing installation from a generated lockfile without running the conda solver. It accepts environment.yml, conda-build meta.yaml, and pyproject.toml dependency specifications. Its default output is a unified lockfile for multiple platforms, which can also be rendered into other lockfile formats. Example targets include Linux, Windows, macOS, Apple silicon, and other Linux architectures. The project documents use alongside conda, mamba, pip, Docker, and pixi, with installation options including pip, conda, and mamba. For isolated installs, pipx or condax are recommended to reduce dependency-conflict risk. Channel authentication supports tokens and basic authentication, with environment-variable substitution for credentials. Mixed conda and pip locking is experimental. Private pip repository support is limited to legacy PyPI repositories with basic authentication.
Who it suits
It suits developers and teams who need repeatable conda environments across multiple operating systems. Users combining pip and conda dependencies should note that mixed-environment locking is experimental.
Good to know, and what to watch
- Creates unified multi-platform lockfiles
- Accepts environment.yml, meta.yaml, and pyproject.toml
- Supports token and basic channel authentication
- Free and MIT licensed
- Mixed conda and pip locking is experimental
- Private pip repositories are limited to legacy PyPI with basic authentication
- Explicit lockfiles with pip packages can only be installed by conda-lock
Plans
Every plan and priceFree and open source
Free plan
The verdict
conda-lock is focused on producing consistent environment lockfiles across target platforms. Its experimental pip support and private repository limit matter for some workflows.
What the maker publishes
- Authentication
- It supports token-based and basic authentication for channels and documents environment-variable substitution for credentials. source
- Credential handling
- The unified lockfile omits raw credential-bearing channel URLs when it can match credential portions to environment variables, while rendered explicit lockfiles may contain substituted variables. source
- Ecosystem integrations
- The documentation describes usage with conda, mamba, pip, Docker, and pixi. source
- Explicit lockfile limitation
- When pip-solved packages are present in an explicit lockfile, the docs say those packages can only be installed by conda-lock. source
- How it works
- It solves dependencies for each target platform and lets installations from generated lockfiles avoid invoking the conda solver. source
- Input formats
- It supports environment.yml, conda-build meta.yaml, and pyproject.toml dependency specifications. source
- Installation
- The project documents installation through pip, conda, and mamba, and recommends isolated installs with pipx or condax to reduce dependency-conflict risk. source
- License
- The repository lists the MIT License and states that the software is provided without warranty. source
- Pip integration
- It has experimental support for locking mixed conda and pip environments using a vendored copy of Poetry’s dependency solver. source
- Private PyPI limit
- The README says private pip repository support is limited to legacy PyPI repositories with basic authentication. source
- Purpose
- conda-lock is a lightweight library for generating fully reproducible lockfiles for conda environments. source
- Support and contributions
- The README directs contributors to its contributing guide for development setup; the documentation links to the GitHub repository. source
- Supported target platforms
- The README lists linux-64, osx-64, win-64, osx-arm64, linux-aarch64, and linux-ppc64le as target platform examples. source
- Unified lockfile
- Its default output is a unified multi-platform lockfile that can be rendered into other lockfile formats. source
Checked against the maker's pages on 4 Oct 2026.