We present a first search algorithm for solving decentralized partially-observable Markov decision problems (DEC-POMDPs) with infinite horizon. The algorithm is suitable for computing optimal controllers for a cooperative group of agents that operate in a stochastic environment such as multi-robot coordination or network traffic control. Solving such problems effectively is a major challenge in the area of planning under uncertainty. Our solution is based on a synthesis of classical best-first search techniques and decentralized control theory. We believe it to be the first optimal search algorithm for this kind of problems, and we present some experimental results on a simple multi-agent coordination task.