An anthropomorfic finger with a transmission system based on tendons has been proposed. This system is able to work in an agonist/antagonist mode. The main problem to control tendons proceeds from the different dimensions between the joint and tendon spaces. In order to solve this problem we propose a position controller that provides motor torques instead of joint torques as proposed in the literature. This position controller is built as a parametric neural network by using of basis functions obtained from the finger structure. This controller insure that the tracking error converges to zero and the weights of the network are bounded. Both control and weight updating has been designed by means of a Lyapunov energy function. In order to improve the computational efficient of the neural network, this has been split up into subnets to compensate inertial, coriolis/centrifugal and gravitational effects.