In the highlands of Madagascar, agricultural expansion gained on grasslands, and cropping systems based on direct seeding with permanent vegetation cover are emerging as a means to sustain upland crop production. The objective of this study was to examine how such agricultural practices affect greenhouse-gas emissions from a loamy Ferralsol previously used as a pasture. We conducted an experiment under controlled laboratory conditions combining cattle manure, crop residues (rice straw), and mineral fertilizers (urea plus NPK or di-NH4-phosphate) to mimic on-field inputs and examined soil CO2 and N2O emissions during a 28-d incubation at low and high water-filled pore space (40% and 90% WFPS). Emissions of N2O from the control soil, i.e., soil receiving no input, were extremely small (< 5 ng N2O-N (g soil)–1 h–1) even under anaerobic conditions. Soil moisture did not affect the order of magnitude of CO2 emissions while N2O fluxes were up to 46 times larger at high soil WFPS, indicating the potential influence of denitrification under these conditions. Both CO2 and N2O emissions were affected by treatments, incubation time, and their interactions. Crop-residue application resulted in larger fluxes of CO2 but reduced N2O emissions probably due to N immobilization. The use of di-NH4-phosphate was a better option than NPK to reduce N2O emissions without increasing CO2 fluxes when soil received mineral fertilizers. Further studies are needed to translate the findings to field conditions and relate greenhouse-gas budgets to crop production.