We propose methods for studying behavior and neural dynamics in deep RL agents based on neuroscience methods and demonstrate them in a naturalistic simulated foraging environment