The human hand is a complex and versatile effector that mediates most of our interactions with the environment. The objective of my dissertation is to shed light on the neural mechanisms that mediate manual behavior in human and non-human primates, focusing on three aspects of prehension: postural control, grasp force control, and force feedback. First, we compare neural population responses in motor cortex during grasp to that of reach and find that neural dynamics associated with grasp are fundamentally different from those associated with reach. During grasp, unlike reach, population responses do not exhibit smooth linear dynamics. Second, we investigate how hand kinematics and interaction forces are encoded in the motor cortex of macaques and develop decoders that accurately track these behavioral variables. We find that linear models can account for the relationship between neural activity in motor cortex and hand posture but not grasp force. Rather, force signals are weak and more dynamic than are postural ones. To capture these patterns requires more complex models that can exploit the force-related dynamics.Next, we use insights gleaned from able bodied animals to develop real-time kinematic and force decoders for individuals with tetraplegia. We confirm that, while kinematics can be decoded linearly from motor cortex, grasp force is best decoded by incorporating non-linear dynamics in human motor cortex, consistent with our results in monkeys. Finally, we develop approaches to incorporate biomimetic sensory feedback into brain computer interfaces aimed at restoring manual dexterity. To this end, we develop a sensory encoding model – which converts the outputs of sensors on the bionic hand into regimes of electrical stimulation applied to the nerve – designed to mimic biological responses and demonstrate that the resulting feedback is more natural and intuitive. The same algorithm can be seamlessly applied to sensory feedback via interfaces with the somatosensory cortex. These results pave the way toward biologically inspired brain-computer interfaces that mimic the functionality of biological hands.