arXiv:2610.41530v1 Announce Type: new Abstract: Neuromorphic controllers interact with physical systems through fixed-amplitude spikes whose timing conveys control information. Existing Lyapunov-based greedy spiking controllers guarantee stability but are restricted to linear systems and neglect firing-rate limits. We extend greedy spiking control to nonlinear systems using Koopman operator theory. For observable dictionaries invariant under differentiation along the input direction, spike-induced state translations admit an exact linear representation in lifted coordinates, enabling exact modeling of spiking jump dynamics. This allows firing decisions to be formulated as comparisons of quadratic Lyapunov functions in the lifted space. We prove practical stability of a clocked, rate-limited controller with an explicit ultimate bound and develop a data-driven design based on Koopman eigenfunctions learned from unforced data by generator-based regression. The weights of the Lyapunov function are synthesized by a linear program that enforces a sampled relaxation of the rate-limited control-Lyapunov condition. Applied to an inverted pendulum, the learned eigenfunctions recover the system energy, and the proposed controller achieves swing-up and stabilization from nearly all initial conditions despite severe actuation limits.