While international cooperation research emphasizes institutional design, states mostly interact with existing organizations. How do states choose organizations for cooperation? We develop a theory of agency choice for development projects, emphasizing the importance of domestic institutions, the scope of cooperation, and the resources of the implementing agency. If states are to cooperate with funding agencies that have abundant resources, such as the World Bank, they must accept more stringent conditions on project implementation. We argue states accept the stringent conditions that resourceful organizations demand if the public goods from project implementation are highly valuable. Empirically, this is the case for democratic states, large projects, and projects that produce national instead of global public goods. We test this theory using data on 2,882 Global Environment Facility (GEF) projects, 1991-2011. The GEF offers an ideal case because various implementing agencies are responsible for the actual projects. States implement projects in collaboration with the World Bank, which has the most expertise and resources among the GEF's implementing agencies, if their regime type is democracy, the project size is large, and the benefits are primarily national. Qualitative evidence sheds light on causal mechanisms.