Robot teleoperation is a transformative field that can enable workers to safely perform tasks in dangerous environments. In this letter, we present our work towards a teleoperation system with safe, realistic force feedback for intuitive control of a robotic arm and anthropomorphic robotic hand as its end effector. The system interfaces with the user via a novel data glove, which detects the state of the hand using inertial measurement units (IMUs) and custom curvature sensors, and employs pneumatic muscles to provide force feedback. We use this glove to control a Kinova Jaco robotic arm and a custom 3D printed hand with embedded force sensors. We tested the functionality of this system in a grasp quality experiment and a full teleoperation test. With haptic feedback, users were accurately able to differentiate between secure and insecure grasps. In user testing with the full system, all users were able to complete a series of ten pick-and-place tasks. Inexperienced users with only 5 minutes of training completed all tasks in an average of 55.1 seconds, while experienced users with 0.5 to 2 hours of practice took an average of 34.6 seconds.