Nitrogen (N) is often the most limiting nutrient to productivity in smallholder mixed crop-livestock systems such as commonly found in the mid-hills and lowland (Terai) of Nepal. Identifying current bottlenecks constraining agroecosystem functioning in terms of N flows and associated improvement options in these systems is paramount. Here, we explore variations in robustness, a concept from ecological network analysis (ENA) which represents the balance of system's degree of order between organization (order/constraint) and adaptive flexibility (freedom/resilience) of N flows. Robustness can provide a detailed assessment of N flows and assist in evaluation of measures to reduce nutrient losses. In this study, the FarmDESIGN model was employed to quantify nitrogen flows, generate ENA indicators of integration, diversity and robustness, and to explore the impact of crop intensification options on N networks across farm types in the mid-hills and lowland (Terai) of Nepal. Results revealed that the farms in the different agroecosystems recycled only a small portion of the total N inputs (<15%), and had therefore high rates of N losses (63–1135¿kg N per ha per year) and high dependency on N imports in the form of fodder (feed self-reliance 11–43%). The farm N networks were organised (high productivity) but inflexible (poorly resilient) and consequently unbalanced (low robustness). Scenarios of improved management (improved seed, intercropping, use of fertilizers, better timing of activities) resulted in improved crop production, leading to reduced fodder imports and less N losses. Consequently, the N networks increased in flexibility which resulted in greater robustness of the N flow network in the farm systems. Increasing on-farm biomass production by improved farm management could be an important element on the way to sustainably intensify smallholder farms, especially when dependency on external resources can be reduced. We conclude that a detailed analysis of nutrient