A Markov Decision Process (MDP) formalizes a sequential decision-making framework with states, actions, and rewards. It assumes the Markov property, where transitions depend only on the current state and action. MDPs are the basis for many reinforcement learning algorithms.