Connect MATLAB R2020a to Python on Ubuntu

MATLAB can be controlled from Python without starting a new MATLAB process for every command. The MATLAB Engine API can attach Python to a shared MATLAB session, allowing a human to keep using the desktop while Python runs commands, scripts, and experiments.

This is especially useful for an older MATLAB release on a current Ubuntu system. MATLAB R2020a supports Python 2.7, 3.6, and 3.7 for its Engine API, while a modern Ubuntu installation may default to a much newer Python version.

• The Target Setup
• Install Python 3.7
• Create an Isolated Environment
• Work Around the R2020a Packaging Metadata
• Install the MATLAB Engine API
• Share the Existing MATLAB Session
• Connect from Python
• Run Commands and Scripts
• Troubleshooting Checklist

The Target Setup

The final connection looks like this:

MATLAB desktop session
        ↕
MATLAB Engine shared session
        ↕
Python client

The Python client connects to the existing MATLAB process. It does not call start_matlab(), which would create another MATLAB instance.

Install Python 3.7

On Ubuntu 24.04, Python 3.7 is not normally available from the standard repositories. The Deadsnakes PPA provides older Python releases:

sudo apt update
sudo apt install -y software-properties-common

sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt update

sudo apt install -y python3.7 python3.7-distutils virtualenv

Python 3.7 is end-of-life, so use it only in an isolated environment for this legacy MATLAB integration. Do not replace Ubuntu’s system Python.

Create an Isolated Environment

Create a dedicated environment with Python 3.7:

mkdir -p /path/to/pyenvs
virtualenv -p /usr/bin/python3.7 /path/to/pyenvs/matlab
source /path/to/pyenvs/matlab/bin/activate

Verify the interpreter:

python --version
which python

The interpreter should be Python 3.7, and the path should point into the new environment.

Work Around the R2020a Packaging Metadata

MATLAB R2020a’s Engine installer is an old setup.py package. Its metadata declares the version as R2020a, which is not a valid PEP 440 version for newer setuptools releases.

Without adjustment, installation can fail with:

InvalidVersion: Invalid version: 'R2020a'

Pin setuptools to an older release inside the MATLAB-specific environment:

python -m pip install --force-reinstall "setuptools==59.6.0"

Do not downgrade setuptools in unrelated Python environments.

Install the MATLAB Engine API

Install the Engine API shipped with the MATLAB installation. Replace the example MATLAB root with the actual installation directory:

python -m pip install \
  /opt/MATLAB/R2020a/extern/engines/python

If the pip-based installation still invokes incompatible modern packaging behavior, use the legacy installer after pinning setuptools:

cd /opt/MATLAB/R2020a/extern/engines/python
python setup.py install

The deprecation warnings emitted by setup.py install are expected for this old package. The installation is successful if the command finishes without an exception.

Test the import:

python -c "import matlab.engine; print('MATLAB Engine imported')"

Share the Existing MATLAB Session

Open the MATLAB desktop and run this command in its Command Window:

matlab.engine.shareEngine('my_matlab')

The name is arbitrary, but it must be used consistently by Python. MATLAB sessions are not shared by default.

Connect from Python

With the virtual environment active, discover shared sessions:

python -c "import matlab.engine; print(matlab.engine.find_matlab())"

The result should contain the session name:

('my_matlab',)

Connect to that existing session:

import matlab.engine

eng = matlab.engine.connect_matlab("my_matlab")

print("Connected")
print("sqrt(16) =", eng.sqrt(16.0))
print("MATLAB folder =", eng.pwd())

connect_matlab() attaches to the desktop session. start_matlab() has different behavior: it launches a new MATLAB process.

Run Commands and Scripts

MATLAB commands that do not return a value should specify nargout=0:

eng.eval("xx = 1;", nargout=0)
print(eng.workspace["xx"])

Calling eng.eval("xx=1") without nargout=0 can produce a MATLAB syntax error because the assignment statement is being requested as though it should return an output.

Run an .m file in the shared session:

eng.cd("/path/to/project", nargout=0)
eng.run("analysis.m", nargout=0)

Retrieve a value from the MATLAB workspace:

value = eng.eval("xx", nargout=1)
print(value)

You can also assign simple values directly:

eng.workspace["gain"] = 2.5
print(eng.workspace["gain"])

Keep the Python process alive if it will issue multiple commands. Reusing the same eng object avoids MATLAB startup time. Avoid calling start_matlab() for each operation, and do not call eng.quit() unless closing the MATLAB session is intentional.

Troubleshooting Checklist

• If the installer rejects Python 3.12, activate the Python 3.7 environment before installing the Engine API.
• If installation fails with InvalidVersion: 'R2020a', pin setuptools to 59.6.0.
• If find_matlab() returns an empty tuple, run matlab.engine.shareEngine('my_matlab') in the desktop MATLAB session.
• Run MATLAB and Python as the same operating-system user.
• Use nargout=0 for assignments, scripts, and commands that do not return values.

The MATLAB Engine API documentation describes the supported Python versions, shared-session functions, workspace access, and MATLAB function calls.