Do multilateral foreign aid institutions allocate their resources in ways that are more consistent with their stated mandates or rather with their donors’ strategic interests to trade aid for influence and policy concessions? In this paper, I argue that dynamics between donors (principals) and institutions (agents) render multilateral aid less prone to capture by principals’ strategic interests than most literature suggests. High monitoring costs, agents’ preferences for survival, and principals’ actions to secure agent survival especially since the end of the Cold War underpin my argument. To support my argument about agent autonomy, I leverage new data on how staff at the World Bank, African Development Bank, Asian Development Bank, and Inter-American Development Bank rank the institutional environments of recipient countries. These data provide a crucial test for my argument because these international organizations purportedly use these institutional ratings data to determine the creditworthiness of recipient countries. Using numerous model specifications relating to the number of loans allocated and money received, I find broad support for my argument on agent autonomy. Especially since the end of the Cold War, countries with positions of power in the international system are less able to pick the winners and losers of development through multilateral aid than most literature suggests. The results of this study contribute to emerging literatures showing that ``aid is not oil’’, that multilateral aid is less prone to capture than bilateral aid, and that bureaucratic factors are crucial for understanding international organizations and foreign aid.