The evolutionary pressures that drive long larval planktonic durations in some coastal marine organisms, while allowing direct development in others, have been vigorously debated. We introduce into the argument the asymmetric dispersal of larvae by coastal currents and find that the strength of the currents helps determine which dispersal strategies are evolutionarily stable. In a spatially and temporally uniform coastal ocean of finite extent, direct development is always evolutionarily stable. For passively drifting larvae, long planktonic durations are stable when the ratio of mean to fluctuating currents is small and the rate at which larvae increase in size in the plankton is greater than the mortality rate (both in units of per time). However, larval behavior that reduces downstream larval dispersal for a given time in plankton will be selected for, consistent with widespread observations of behaviors that reduce dispersal of marine larvae. Larvae with long planktonic durations are shown to be favored not for the additional dispersal they allow, but for the additional fecundity that larval feeding in the plankton enables. We analyzed the spatial distribution of larval life histories in a large database of coastal marine benthic invertebrates and documented a link between ocean circulation and the frequency of planktotrophy in the coastal ocean. The spatial variation in the frequency of species with planktotrophic larvae is largely consistent with our theory; increases in mean currents lead to a decrease in the fraction of species with planktotrophic larvae over a broad range of temperatures.