Many cellular processes require telling time. Cells may use kinetic filtering, the ability to respond to sustained but not transient stimulation, to decode information stored in dynamical profiles and exert control over relative timing of events. Previous work has found only one circuit architecture, the AND-gated coherent feed forward loop, capable of such behavior. We enumerate the space of 1-, 2-, and 3-node networks to ask what additional architectures can drive kinetic filtering, which is quantified by measuring temporal dose response steepness and trigger time. We find two types of positive feedback architectures, two types of buffering architectures, and the coherent feed forward loop with AND logic are the only kinetic filters of 3 or fewer components. The buffering double inhibition circuit architecture is modular, reversible, capable of the widest range of trigger times, and most suitable as a template for future engineering efforts.