mpcrl is a library that combines Model Predictive Control with Reinforcement Learning, using MPC as a function approximator to train RL agents. It uses CasADi for optimization and supports Python 3.11+. The library provides tools for defining RL agents and training them with model-based predictive control strategies.