Current international relations theory struggles to explain both the autonomy and transformation of international organizations (IOs). Previous theories either fail to account for any IO behavior that deviates from the interests of member states, or neglect the role of member states in reforming IO institutions and behavior. We propose an agency theory of IOs that can fill these gaps while also addressing two persistent problems in the study of IOs: common agency and long delegation chains. Our model explains slippage between member states' interests and IO behavior, but also suggests institutional mechanisms—staff selection, monitoring, procedural checks, and contracts—through which states can rein in errant IOs. We evaluate this argument by examining multiple institutional reforms and lending patterns at the World Bank from 1980 to 2000.