Scholars have long understood that the instability of power has ramifications for compliance with international law. Scholars have not, however, focused on how states’ expectations about shifting power affect the initial design of international agreements. In this paper, I integrate shifting power into an analysis of the initial design of both the formal and substantive aspects of agreements. I argue that a state expecting to become more powerful over time incurs an opportunity cost by agreeing to formal provisions that raise the cost of exiting an agreement. Exit costs—which promote the stability of legal rules—have distributional implications. Before joining an agreement, an “ascendant” state will therefore often require either a greater share of the benefits from cooperation, or a reduction in exit costs through the use of formal provisions such as withdrawal clauses, sunset clauses, and provisions affecting the legality of an agreement. I analyze how states determine which concessions to make in order to reach agreement with an ascendant state. This analysis helps explain a number of puzzles in the international legal literature, such as why states with poor compliance rates are sometimes observed to join international agreements at the same or higher rates than states with good compliance rates; why weak agreements often evolve into more constraining agreements; and why multilateral agreements are more likely to have low exit costs than bilateral agreements.