Leptospirosis is a neglected zoonotic disease and one of the leading causes of zoonotic morbidity and mortality, particularly in resource-poor settings. Sri Lanka has one of the highest disease burdens worldwide, with occasional endemic leptospirosis outbreaks (2008, 2011). Rodents are considered the main wildlife reservoir, but due to a scarcity of studies it is unclear which particular species contributes to bacterial transmission and reservoir maintenance in this multi-host multi-parasite system. Several rodent species act as agricultural pests both in rice fields and in food storage facilities. To unravel the interactions among the small mammal communities, pathogenic Leptospira spp. and human transmission pathways, we collected animals from smallholder food storage facilities, where contact between humans and small mammals is most likely, and screened kidney tissue samples for Leptospira spp. using PCR. Samples were collected in three climatic zones along a rainfall gradient. Pathogenic Leptospira spp. were detected in small mammal communities in 37 (74%) out of 50 sampled farms and 61 (12%) out of 500 collected individuals were infected. The small mammal community was comprised of Rattus rattus (87.6%), Suncus shrews (8.8%), Bandicota spp. (2.8%) and Mus booduga (0.8%). Three pathogenic Leptospira spp. were identified, L. borgpetersenii (n = 34), L. interrogans (n = 15), and L. kirschneri (n = 1). Suncus shrews were commonly infected (32%), followed by B. indica (23%) and R. rattus (10%). L. borgpetersenii strains similar to strains previously extracted from human clinal samples in Sri Lanka were detected in R. rattus and Suncus shrews. L. interrogans was observed in R. rattus only. A single L. kirschneri infection was found in M. booduga. The presence of human pathogenic Leptospira species in an agricultural pest rodent (R. rattus) and in commensal shrews (Suncus) calls for management of these species in commensal settings. Further investigation of the interplay between pathogen and reservoir population dynamics, overlap in geographic range and the extent of spill-over to humans in and around rural settlements is required to identify optimal management approaches.