This paper deals with an application of an adaptive control to a magnetically levitated microrbot. Using an electromagnetic device, a microrobot is levitated and manipulated within a 3-D space inside a magnetic field. The microrobot has two fingers that can grasp and elevate objects. PID controls are applied for positioning of the microrobot in the three axes. However, as the microrobot deals with various payloads, a PID control may not be sufficient to maintain the microrobot on high performance in the vertical axis. To improve the performance, an adaptive control law is also examined for the positioning in the vertical axis so that the controller parameters become adjustable in real-time to cope with uncertainties and variations in payloads. A model-reference adaptive system (MRAS) based on the augmented error is designed, and simulations and experiments are conducted to verify the effectiveness of the control. The simulation and experimental results for PID and adaptive control are shown for comparison.