HILO-MPC is a Python toolbox for machine-learning-supported optimal control and estimation, developed at TU Darmstadt and Otto von Guericke University. It provides a comprehensive set of controllers (NMPC, LMPC, PID, LQR), state estimators (MHE, Kalman filters, particle filter), and machine learning models (Gaussian processes, neural networks) that can be integrated into control loops. The architecture is built on CasADi for symbolic differentiation and optimization, with a plugin system for backends (PyTorch, TensorFlow, scikit-learn) and visualization (Matplotlib, Bokeh). The toolbox supports ODE/DAE modelling, linearization, discretization, and code generation for embedded deployment. The codebase is organized into modules for controllers, estimators, dynamic models, machine learning, and utilities, with a clear base class hierarchy. Key reusable components include symbolic variable management (Vector, RightHandSide), hyperparameter optimization frameworks, kernel composition for GPs, and OPC UA connectivity for industrial control. The project includes extensive documentation, examples, and tests covering chemical reactors, bioreactors, robotics, and automotive applications. While primarily an application library for control engineering, it also incorporates research-oriented features such as stochastic MPC, path following, and hybrid modelling with learned components. The code is licensed under LGPL-3.0 and is actively maintained with CI/CD pipelines and regular releases.