Blender ships with its own embedded version of Python. Therefore, the simplest (and recommended) way to install COMPAS for Blender is to replace the embedded Python by the Python of a conda environment that already has COMPAS installed.

It is important that the version of Python installed in the conda environment matches the version of Python that was originally shipped with Blender. For Blender 2.8x and 2.9x this is Python 3.7x.


If you don’t have an environment yet with Python 3.7 and COMPAS you can create one with conda.

conda config --add channels conda-forge
conda create -n blender python=3.7 COMPAS

Configuring Blender to use the newly installed environment is slightly different per OS.

conda activate blender
python -m compas_blender.install "%PROGRAMFILES%\\Blender Foundation\\Blender 2.91\\2.91"

Note that the path %PROGRAMFILES%\\Blender Foundation\\Blender 2.91\\2.91 might be different if you have another version of Blender intalled. Check your version of Blender and change the path accordingly.

conda activate blender
python -m compas_blender.install /Applications/

Note that the path /Applications/ might be different if you have another version of Blender intalled. Check your version of Blender and change the path accordingly.

conda activate blender
python -m compas_blender.install ~/Blender/2.91

Note that the path ~/Blender/2.91 might be different for your setup.


For some Blender add-ons, not only the version of Python has to match, but also the version of Numpy. For example, Sverchok, a Grasshopper-type visual programming tool for Blender, will not work with the version of Numpy included in the latest COMPAS releases, even though Blender will.

In those cases, you can simply revert to an earlier version of Numpy that is still compatible with COMPAS in the environment you use with Blender. For Sverchok, this would be version 1.17.5, which is the one shipped with Blender originally. To revert simply do

conda activate blender
conda install numpy=1.17.5

Verify setup

In Blender, at the interactive Python prompt (>>>) import the following packages

>>> import compas
>>> import compas_blender
>>> import numpy
>>> import scipy

Install Python packages

After replacing the embedded version of Python with the one in the virutual environment, as instructed above, it is not necessary to activate the environment before using Blender. However, when you want to update the installed packages, or add more packages, make sure to activate the Blender environment first. Otherwise, the changes will not have any effect.

conda activate blender
conda install ...


If you want to use blender with a different environment, you simply have to activate that environment and follow the same procedure described above.

Start Blender

Blender uses the command window of your system to display error messages and other text output. On Windows, the output window can be brought to the front directly from the “Window” menu. On Mac or Linux, you should start Blender from the command line.

By adding the Blender executable to the PATH variable this is really simple. Just add the following to your .bash_profile or .bashrc.

export PATH="/Applications/$PATH"
export PATH="~/Blender/2.91:$PATH"

Note that this path might be different on your system.

After that starting Blender from the command line is much simpler.


Scripting Interface

To switch to the scripting interface, simply select the “Scripting” tab of the main window.

The scripting interface has an embedded interactive Python terminal, which is located in the bottom half of the main window on the left. If COMPAS was successfully installed you can use it to directly access the conda environment from where the installation was executed.

>>> import compas
>>> import compas_blender
>>> import numpy
>>> import scipy
>>> import bpy

The script editor is quite simple but good enough for basic development. Line numbers and syntax highlighting should be on by default, but if that is not the case, they can be turned on with toggle buttons at the top right of the area. Further customisation of the editor appearance is possible by opening the sidebar from the “View” menu of the editor.

Basic Usage

One of the main advantages of working in Blender is that Blender Python is CPython, and not IronPython like in Rhino and Grasshopper. This means that all cool Python libraries are directly available and do not need to be accessed through remote procedure calls (RPC). Especially for code that relies heavily on libraries such as Numpy and Scipy this simplifies the development process quite significantly.

import compas
import compas_blender
from compas.datastructures import Mesh
from compas_blender.artists import MeshArtist


mesh = Mesh.from_ply(compas.get('bunny.ply'))

artist = MeshArtist(mesh)

Data Blocks

Something worth explaining is the use of compas_blender.clear() in this script. Blender uses (and re-uses) something called “data blocks”. Objects in the scene have instances of these data blocks assigned to them. Multiple objects can be linked to the same data block. As a result, simply deleting an object from the scene will delete the object but not the underlying data block.

If you run a script multiple times, even if you delete the scene objects between consecutive runs, you will accumulate the data blocks from previous runs and after a while Blender will become very slow.

compas_blender.clear() attempts to clean up not only the scene objects but also the data blocks. If somehow you still experience a slowdown, restarting Blender will help (all unused data blocks are then automatically removed).


There are no real layers in Blender; at least not like the layers in, for example, Rhino. Therefore, the Blender artists have no optional layer parameter and no clear_layer method. Instead, objects are grouped in collections, which can be turned on and off in the Blender UI similar to layers in Rhino.



compas_blender is not yet as well developed as, compas_rhino and compas_ghpython. For example, COMPAS geometry objects do not yet have a corresponding artist in compas_blender. Artists are currently only available for data structures and robots.

There is also no official system yet for making custom COMPAS tools in Blender. Therefore, COMPAS Blender development is somewhat limited to individual scripts.