A new rehabilitation robot is presented which targets acute stroke patients. Observational fieldwork on the stroke ward at a local hospital informed the initial design, a linear 1-DOF robot targeting the knee flexion/extension exercise typical to traditional bed-bound rehabilitation. An admittance controller was developed to apply assistive or resistive forces to the patient's leg. Three games were created which make the exercise more engaging, along with haptic feedback control which can render virtual forces to the patient through the robot. We conducted an experiment on healthy subjects, which looked to evaluate the effects of assistance and resistance levels, along with the participant's view on the device through a subjective questionnaire. The experiments indicate that assistance does work to help the user follow the desired trajectory, that resistance can be used to make the task more difficult, and that the games have potential for making the exercise more engaging.