We study the optimality conditions of information transfer in systems with memory in the low signal-to-noise ratio regime of vanishing input amplitude. We find that the optimal mutual information is represented by a maximum variance of the signal time course, with correlation structure determined by the Fisher information matrix. We provide illustration of the method on a simple biologically inspired model of electrosensory neuron. Our general results apply also to the study of information transfer in single neurons subject to weak stimulation, with implications to the problem of coding efficiency in biological systems.