Rapid electric-vehicle (EV) uptake in Kenya and comparable emerging markets is placing spatially concentrated loads on radial feeders that were not designed for them. A low-carbon national generation mix does not guarantee that an individual feeder has the voltage headroom to absorb clustered charging. This paper formulates charging-station siting and sizing as a constrained, mixed discrete–continuous, voltage-stability-aware planning problem and solves it using the artificial bee colony (ABC) algorithm with a backward–forward sweep load flow embedded in the fitness evaluation. A branch-based radial stability index that depends explicitly on active power is adopted and enforced as a constraint during the search, because the reactive-power-driven indices in common use respond only weakly to unity-power-factor charging. Uncoordinated and optimized deployments are compared on the IEEE 33-bus feeder at three penetration levels defined by installable charger counts, and EV hosting capacity is quantified under both statutory and operational voltage limits. Optimized siting reduces active-power losses by 30.3%, raises the minimum bus voltage by 0.0157 p.u. and increases hosting capacity by 247% relative to uncoordinated deployment, with results reported over 30 independent runs.