This paper presents a method for three-dimensional attitude stabilization of a satellite. The pitch loop of the satellite is controlled by a momentum wheel; whereas the roll/yaw loops are stabilized using two magnetic torques along their respective axes. In order to design an efficient controller, the stability conditions are considered based on a nonlinear model of system. An adjustable adaptive fuzzy system is proposed as the method to design the controller. The span of membership functions are tuned using errors of fuzzy inputs with respect to their references. Results show that fuzzy sets cover all variations of fuzzy inputs and optimal fuzzy output is gained. The Lyapunov synthesis method is used to prove the stability of the closed-loop system. The efficiency of the controller in converging of the position error to close to zero is also shown using some numerical simulations.