Osteoclasts are multinucleated giant cells that resorb bone, ensuring development and continuous remodelling of the skeleton and the bone marrow haematopoietic niche. Defective osteoclast activity leads to osteopetrosis and bone marrow failure1-9, whereas excess activity can contribute to bone loss and osteoporosis10. Osteopetrosis can be partially treated by bone marrow transplantation in humans and mice11-18, consistent with a haematopoietic origin of osteoclasts13,16,19 and studies that suggest that they develop by fusion of monocytic precursors derived from haematopoietic stem cells in the presence of CSF1 and RANK ligand1,20. However, the developmental origin and lifespan of osteoclasts, and the mechanisms that ensure maintenance of osteoclast function throughout life in vivo remain largely unexplored. Here we report that osteoclasts that colonize fetal ossification centres originate from embryonic erythro-myeloid progenitors21,22. These erythro-myeloid progenitor-derived osteoclasts are required for normal bone development and tooth eruption. Yet, timely transfusion of haematopoietic-stem-cell-derived monocytic cells in newborn mice is sufficient to rescue bone development in early-onset autosomal recessive osteopetrosis. We also found that the postnatal maintenance of osteoclasts, bone mass and the bone marrow cavity involve iterative fusion of circulating blood monocytic cells with long-lived osteoclast syncytia. As a consequence, parabiosis or transfusion of monocytic cells results in long-term gene transfer in osteoclasts in the absence of haematopoietic-stem-cell chimerism, and can rescue an adult-onset osteopetrotic phenotype caused by cathepsin K deficiency23,24. In sum, our results identify the developmental origin of osteoclasts and a mechanism that controls their maintenance in bones after birth. These data suggest strategies to rescue osteoclast deficiency in osteopetrosis and to modulate osteoclast activity in vivo.