Abstract Biomass is a key variable for crop monitoring and for assessing carbon stocks and bioenergy potential. This study aimed to develop an allometric model for predicting the dry leaf biomass of sisal, an agave plant with crassulacean acid metabolism grown for fibre production in the tropics and subtropics and whose biomass can be utilised as a feedstock to produce biogas through anaerobic digestion. The allometric model was used to estimate leaf biomass and productivity across different stand ages in a sisal plantation in semi-arid region in south-east Kenya (annual rainfall 611 mm and temperature 24.9 °C). Based on a sample of 38 leaves, the best predictor for biomass was leaf maximum width and plant height used as a combined variable in a log-log regression model (cross-validated R2 = 0.96 and root-mean-square error = 7.69 g). The mean productivity in nine 26- to 36-month-old plots was 11.1 Mg ha−1 yr−1, which could potentially yield approximately 3000 m3 CH4 ha−1 yr−1. The leaf biomass in 55 field plots (400 m2 in area) ranged from 2.7 to 42.7 Mg ha−1, with mean at 13.5 Mg ha−1, which equals to 6.3 Mg C ha−1. The yielded allometric equations can be utilised for predicting the leaf biomass of sisal in similar agro-ecological zones. The estimates on plantation biomass can be used in assessing the role of sisal plantations as a regional carbon storage. In addition, the results provide reference on the productivity of agave and crassulacean acid metabolism in semi-arid regions of East Africa, where such reports are few.