Electrical vagus nerve stimulation is a treatment alternative for many epileptic and depressed patients whose symptoms are not well managed with pharmaceutical therapy. However, the fixed stimulus, open loop dosing mechanism limits its efficacy and precludes major advances in the quality of therapy. A real-time, responsive form of vagus nerve stimulation is needed to control nerve activation according to therapeutic need. This personalized approach to therapy will improve efficacy and reduce the number and severity of side effects. We present autonomous neural control, a responsive, biofeedback-driven approach that uses the degree of measured nerve activation to control stimulus delivery. We demonstrate autonomous neural control in rats, showing that it rapidly learns how to most efficiently activate any desired proportion of vagal A, B, and/or C fibers over time. This system will maximize efficacy by minimizing patient response variability and by minimizing therapeutic failures resulting from longitudinal decreases in nerve activation with increasing durations of treatment. The value of autonomous neural control equally applies to other applications of electrical nerve stimulation.