Many food production systems in sub-Saharan Africa are constrained by phosphorus (P). We hypothesized that within legume-cereal rotation systems: targeting P to the legume phase leads to higher system productivity, and that use of grain legumes leads to better economic returns than use of herbaceous legumes. Four P application regimes: (i) no P, (ii) P applied every season, (iii) P applied in season 1 only and (iv) P applied in season 2 only were tested for four seasons in three cropping systems (continuous maize, mucuna-maize rotation and soybean-maize rotation) in a split plot experiment set up in Nyabeda, western Kenya. Treatments where P was applied were better than no P treatments. While continuous cereal systems showed the need for application of P every second season, rotation systems involving mucuna and soyabean indicated that application in one out of three seasons could be sufficient. Nitrogen fertilizer equivalence was 52 to >90 kg N ha−1 for soyabean and 37 to >90 kg N ha−1 for mucuna, depending on P fertilization and season. Analysis of marginal rates of return (MRR) showed that soybean-maize rotation with one application of P was the most economically viable option, with an MRR of at least 147% compared to other non-dominated options.