This paper investigates the adaptive output containment control of general linear heterogeneous multi agent systems with multiple unknown leaders whose dynamics are only known to the neighboring followers. The output containment control objective is to assure the convergence of each follower's output to the dynamic convex hull spanned by the leaders' outputs. Compared with the situation where the leaders' dynamics are known to each follower exactly, local adaptive observers are presented at each follower to estimate the leaders' dynamics. Then, an adaptive tuning law is proposed to solve the associated output regulator equations without knowing leaders' dynamics. Two distributed control protocols, using state-feedback and dynamic output-feedback are locally designed, using the estimated leaders' dynamics and adaptive solutions to the output regulator equations. Both proposed control protocols are verified using numerical simulations. (C) 2018 Elsevier Ltd. All rights reserved.