The eukaryotic cell cycle comprises an ordered series of events, orchestrated by the activity of cyclin-dependent kinases (Cdks), leading from chromosome replication during S phase to their segregation in mitosis. The unidirectionality of cell-cycle transitions is fundamental for the successful completion of this cycle. It is thought that irrevocable proteolytic degradation of key cell-cycle regulators makes cell-cycle transitions irreversible, thereby enforcing directionality. Here we have experimentally examined the contribution of cyclin proteolysis to the irreversibility of mitotic exit, the transition from high mitotic Cdk activity back to low activity in G1. We show that forced cyclin destruction in mitotic budding yeast cells efficiently drives mitotic exit events. However, these remain reversible after termination of cyclin proteolysis, with recovery of the mitotic state and cyclin levels. Mitotic exit becomes irreversible only after longer periods of cyclin degradation, owing to activation of a double-negative feedback loop involving the Cdk inhibitor Sic1 (refs 4, 5). Quantitative modelling suggests that feedback is required to maintain low Cdk activity and to prevent cyclin resynthesis. Our findings demonstrate that the unidirectionality of mitotic exit is not the consequence of proteolysis but of systems-level feedback required to maintain the cell cycle in a new stable state.