Obesity, insulin resistance and increased propensity for type 2 diabetes and cardiovascular disease result from an imbalance between energy intake and expenditure. The cloning of genes involved in energy homeostasis produced a simple feedback model for the homeostatic regulation of adipose mass. Serum leptin secreted from adipocytes signals nutrient sufficiency, curbing appetite and supporting energy expenditure. A rapid decline in leptin during nutrient scarcity instigates adaptive mechanisms, including increased appetite and reduced energy expenditure. Hypothalamic melanocortin neurons are important mediators of this response, integrating inputs of energy status from leptin with other peripheral signals. While this feedback response prolongs survival during fasting, other mechanisms allowing the prediction of nutrient availability also confer a selective advantage. This adaptation has been commonly studied in rodents using restricted feeding paradigms constraining food intake to limited periods at 24-h intervals. Restricted feeding rapidly elicits rhythmic bouts of activity and wakefulness anticipating food presentation. While the response exhibits features suggesting a clock-like mechanism, the neuromolecular mechanisms governing expression of food anticipatory behaviours are poorly understood. Here we discuss a model whereby melanocortin neurons regulating the homeostatic adaptation to variable caloric availability also regulate inputs into neural networks governing anticipatory rhythms in wakefulness, activity and metabolism.