Lower greenhouse gas (GHG) emissions from legume-based cropping systems have encouraged their use to deliver mitigation in agricultural systems. Considerable uncertainties remain about the interaction of legumes with long-term tillage systems on GHG emissions under rainfed agroecosystems. In this context, a field experiment was undertaken under a rainfed vetch crop to evaluate the effect of three long-term tillage systems (i.e. no tillage (NT), minimum tillage (MT) and conventional tillage (CT)) on nitrous oxide (N 2 O) and methane (CH 4 ) emissions for 1 year. Different N 2 O flux patterns were observed among tillage systems during the growth period of vetch, which depended on the soil conditions favouring nitrification and denitrification. The NT system maintained a higher sink for N 2 O than MT and CT from January to mid-April, which significantly reduced N 2 O emissions at this stage. In this period, denitrification capacity and nirK gene numbers were higher for MT than NT and CT. Additionally, an increase in soil NO 3 − content and more favourable denitrification conditions in MT and NT than in CT for the last crop period increased N 2 O emissions in conservation tillage systems. Total annual N 2 O losses were significantly higher in MT (124.2 g N 2 O-N ha −1 ) than NT (51.1 g N 2 O-N ha −1 ) and CT (54 g N 2 O-N ha −1 ) in a vetch crop. Low net uptake of CH 4 was observed for all tillage systems. These results suggested that long-term NT may be a better option than MT to mitigate GHG emissions in rainfed legume-cereal rotation.