Cromwell releases are available at the GitHub Releases page. You are strongly encouraged to use the latest release of Cromwell whenever possible.
Cromwell is distributed as a conda package on conda-forge.
These instructions need to be followed for installing the miniconda distribution and
activating the conda-forge channel. After this Cromwell can be installed in the
base environment with
conda install cromwell or a separate environment for Cromwell can be created with
conda create -n cromwell cromwell. If you are using Cromwell for bioinformatics workflows, you might like to take
a look at bioconda as well.
The conda installation of Cromwell comes with a wrapper that locates the jar for you and allows for running Cromwell or Womtool with a
womtool validate or other command. Conda also installs the required Java dependency
in the environment automatically.
Mac users with Homebrew can also get Cromwell with the command
brew install cromwell.
This documentation frequently refers to a "Cromwell jar" with a name like
This is the main artifact in Cromwell releases that contains all executable Cromwell code and default configuration.
Java 8 is required to run Cromwell.
For users running a Cromwell server a docker image has been made available.
Apple Silicon support statement (updated 2020-11-17)
Cromwell JAR works out of the box
The Cromwell JAR works on any standard Java installation. A user can install an x86 Java runtime on an Apple Silicon Mac and the Rosetta 2 translation layer runs Cromwell at near-native speed.
Once natively-compiled Java runtimes become available, performance will increase with no change in functionality.
Docker Desktop support is in progress
The Cromwell Docker image will not run on M1 Macs until Docker Desktop ships the appropriate update. For more details, please see their official announcement.
By extension, the absence of Docker means that Cromwell's local Docker backend is not yet supported.
Even when Docker Desktop goes native on Apple Silicon, any tool images running on the local backend will need to cross-compile for the x86 and Arm architectures. This is because the Rosetta 2 translation layer does not support virtualization. Please contact the tool maintainers for more information.