In this paper, adaptive finite time consensus control problem is investigated for multiple mechanical systems with uncertainties. New adaptive distributed control algorithms are proposed for the multiple mechanical systems under a switching undirected graph, which is addressed through a novel Markov chain. Without any detailed information of the system model, the adaptive distributed controller has a faster and higher precision tracking performance. It is shown that the states of the mechanical systems can reach a consensus within finite time. And finally, an illustrative simulation on the multiple robotic manipulators is given to demonstrate the effectiveness of the proposed controller.