In this paper, a new model reference adaptive output feedback controller is proposed for general multi-input multi-output (MIMO) systems with unknown parameters such that the tracking errors as well as the control gains remain uniformly bounded. First an adaptive observer is designed to estimate the state variables and system parameters from input-output data. Then a linear combination of the estimated states and the model reference states are used to design a robust adaptive output feedback controller. Several theorems are given to show the convergence of the modeling errors and the gains. Examples are also given to illustrate the effectiveness of the proposed method.