The osteoblast‐derived protein osteocalcin has recently been shown to affect adiposity and glucose homeostasis in mice, suggesting that the skeleton influences energy metabolism through an endocrine mechanism. The aim of this study was to investigate the relationship between plasma osteocalcin and parameters reflecting fat mass and glucose homeostasis in humans. Fasting levels of plasma osteocalcin, plasma glucose, serum insulin, and lipids were analyzed in elderly men (75.3 ± 3.2 yr of age) in the Gothenburg part (all subjects, n = 1010; nondiabetic, n = 857; diabetic, n = 153) of the MrOS Sweden study. Fat mass and lean mass were analyzed using DXA. Diabetic subjects had lower plasma osteocalcin (−21.7%, p < 0.001) than nondiabetic subjects. For both all subjects and nondiabetic subjects, plasma osteocalcin was clearly inversely related to body mass index (BMI), fat mass, and plasma glucose (p < 0.001), whereas it was not associated with height or lean mass. Plasma osteocalcin explained a substantial part (6.3%) of the variance in plasma glucose, whereas it associated moderately with serum insulin. Multiple linear regression models adjusting for serum insulin and fat mass showed that plasma osteocalcin was an independent negative predictor of plasma glucose (p < 0.001). We herein, for the first time in humans, show that plasma osteocalcin is inversely related to fat mass and plasma glucose. Although one should be cautious with mechanistic interpretations of cross‐sectional association studies, our human data support recently published experimental studies, showing endocrine functions of osteoblast‐derived osteocalcin on glucose and fat homeostasis.