The use of anthropomorphic mechanical hands allows the execution of complex manipulation tasks and, therefore, its use is increasing, both in humanoid robots and mobile manipulators. The planning of the motions of these hands is not easy due to the high number of degrees of freedom (DOF) involved. Even when they may be mechanically independent, these degrees of freedom can be artificially coupled in order to mimic the human hand motions by using synergies, which in its turn results in a lower subspace where to perform the motion planning, as previously done by the authors using a PRM. This paper extends this idea by using an RRT*, that differently from other sampling-based planners, is an asymptotically optimal method. The optimization function selected evaluates the alignment of the path to the directions defined by the synergies, thus favoring human-like motions. The proposal is conceptually illustrated with a simple 2 DOF planar robot and applied to a 13 DOF mechanical hand.