We study a nutrition-sensitive agricultural program in low-income rural Zambia between 2011 and 2015. Using a pre-post design with a control group, we measure program effects along established pathways connecting agriculture to nutrition: diversity of agricultural production, crop sales, household food access and child and maternal diets. The program increased diversity in crops grown and the number of months in which various food groups were harvested. In particular, the program substantially increased the percentage of households producing three nutritious crops it promoted (groundnuts, rape and tomatoes). As a consequence there were modest increases in household access to diverse food groups. Despite modest increases in the proportion of children consuming pulses, legumes and nuts, ultimately there were no significant improvements in the overall dietary diversity of young children or their mothers. A nutrition-sensitive agricultural program can increase diversity in agricultural production and to a lesser extent access to nutritious foods, but this may not always be sufficient to improve child diets or nutrition.