This paper describes a numerical algorithm to solve the inverse kinematics of parallel robots based on numerical integration. Inverse kinematics algorithms based on numerical integration involve the drift phenomena of the solution; as a consequence, errors are generated when the end-effector location differs from that desired. The proposed algorithm associates a novel method to describe the differential kinematics with a simple numerical integration method. The methodology is presented in this paper and its exponential stability is proved. A numerical example and a real application are presented to outline its advantages.