This paper deals with the development of an Optical Coherence Tomography (OCT) based visual servoing. The proposed control law uses the wavelet coefficients of the OCT images as the signal control inputs instead of the conventional geometric visual features (points, lines, moments, etc.). An important contribution is the determination of the interaction matrix that links the variation of the wavelet coefficients to the OCT probe (respectively to the robotic platform) spatial velocity. This interaction matrix, required in the visual control law, is obtained from time-derivation of the wavelet coefficients. This work is carried out in a medical context which consists of automatically moving a biological sample in such a way to go back to the position of a sample region that corresponds to a previous optical biopsy (OCT image). For instance, the proposed methodology makes it possible to follow accurately the progress of a pathological tissue between an optical biopsy and a former one. The developed method was experimentally validated using an OCT imaging system placed in an eye-to-hand configuration viewing the robotic platform sample holder. The obtained results demonstrated the feasibility of this type of visual servoing approach and promising performances in terms of convergence and accuracy.