The purpose of our research was to develop a framework for evaluating the sustainability of cropping systems for resource-poor farmers based on local socioeconomic, cultural and biophysical constraints that operate within and between different timeframes. We suggest that the most sustainable cropping system may represent a compromise between factors operating at different timeframes such that farmers obtain short and medium-term incentives to actually adopt cropping systems that offer the longer-term environmental benefits. We then employed the framework to compare the sustainability of three maize/legume relay crops in smallholder farmers' cropping systems in southern Malawi. The three legumes were Sesbania sesban, Tephrosia vogelii, and Cajanus cajan (pigeonpea). We investigated the cropping systems' short-term livelihood impacts by analyzing on-farm ex ante adoption potential and ex post adoption. We investigated the medium-term impacts by comparing the cropping system recommendations based on: (1) maize yields alone, (2) maize yields and production risk (or year-to-year variation in yields), and (3) maize yields, production risk, and farmers' levels of vulnerability. The recommendations were also disaggregated by landscape. Finally, we investigated the long-term timeframe by analyzing the cropping systems impacts on long-term soil quality and maize productivity. We concluded that the pigeonpea/maize relay cropping system was, in general, the most sustainable of the three cropping systems. This was due to the short-term livelihood benefits it provides as a secondary food crop prevalent in the local diet, and our findings that the legumes had relatively equivalent impacts on maize yields and soil quality, although this was dependant upon landscape. However, we also concluded that farmers' unique socioeconomic and agroecological circumstances strongly influenced which cropping system was most sustainable for them. We suggest that participatory investigation of locally relevant cultural, socioeconomic, and agroecological factors vastly improved the quality of our findings and ability to make recommendations to farmers. Basing recommendations on maize yields alone would have potentially placed farmers at greater risk by overlooking important constraints to cropping system adoption. We encourage others to employ and fine-tune the framework we developed for evaluating cropping system sustainability and suggest it could best be carried out by interdisciplinary teams.