Osteoclasts have a unique bone-destroying capacity, playing key roles in steady-state bone remodeling and arthritic bone erosion. Whether the osteoclasts in these different tissue settings arise from the same precursor states of monocytoid cells is presently unknown. Here, we show that osteoclasts in pannus originate exclusively from circulating bone marrow-derived cells and not from locally resident macrophages. We identify murine CX3CR1hiLy6CintF4/80+I-A+/I-E+ macrophages (termed here arthritis-associated osteoclastogenic macrophages (AtoMs)) as the osteoclast precursor-containing population in the inflamed synovium, comprising a subset distinct from conventional osteoclast precursors in homeostatic bone remodeling. Tamoxifen-inducible Foxm1 deletion suppressed the capacity of AtoMs to differentiate into osteoclasts in vitro and in vivo. Furthermore, synovial samples from human patients with rheumatoid arthritis contained CX3CR1+HLA-DRhiCD11c+CD80-CD86+ cells that corresponded to mouse AtoMs, and human osteoclastogenesis was inhibited by the FoxM1 inhibitor thiostrepton, constituting a potential target for rheumatoid arthritis treatment.