Huskarl is a framework for deep reinforcement learning focused on modularity and fast prototyping. It is built on TensorFlow 2.0 using the tf.keras API and enables parallelization of environment dynamics across multiple CPU cores to speed up on-policy learning algorithms. The framework integrates with OpenAI Gym and implements multiple RL algorithms including DQN, Double DQN, Dueling DQN, A2C, DDPG, and Prioritized Experience Replay.