Short-season grain legumes play an important role in smallholder farming systems as source of food and to improve soil fertility through nitrogen fixation. However, it is not clearly understood how these diverse legumes contribute to the resilience of such systems in semi-arid environments. We describe the growth, development and resource-use efficiency (focusing on radiation, RUE) of three promising short-season grain legumes: common bean (Phaseolus vulgaris L.), cowpea (Vigna unguiculata (L.) Walp.) and lablab (Lablab purpureus (L.) Sweet). Two field experiments were conducted during the short rains of 2012–13 and 2013–14 in Eastern Kenya. In the first experiment, the legumes were grown at three plant densities (low, medium, high); in the second experiment, they were subjected to three water regimes (rainfed, partly irrigated, fully irrigated). Phenological development was monitored and biomass accumulation, leaf area index and fractional radiation interception were measured repeatedly during growth; grain yield was measured at maturity. Harvest index and RUE were calculated from these data. Common bean had the shortest growing period (70 days), the most compact growth habit and relatively high RUE but limited grain yield (1000–1900 kg ha^–1), thereby proving more suitable for cultivation in areas with restricted cropping windows or in intercropping systems. Cowpea had a longer growing period (90 days) and a spreading growth habit leading to high light interception and outstanding grain yields under optimal conditions (1400–3050 kg ha^–1). Lablab showed stable RUE values (0.76–0.92 g MJ^–1), was relatively unaffected by limited water availability and had a comparatively long growing period (100 days). Lablab grain yields of ∼1200–2350 kg ha^–1 were obtained across all water regimes, indicating a high potential to cushion climatic variability. Planting density strongly influenced the production success of cowpea and lablab, with high plant densities leading to vigorous growth habit with low podset establishment. Such information on temporal and spatial differences in growth, development and resource-use efficiency is highly valuable for crop-modelling applications and for designing more resilient farming systems with short-season grain legumes.