In this paper, we consider the problem of pursuit-evasion using multiple Autonomous Underwater Vehicles (AUVs) in a 3D water volume, with and without simple obstacles. Pursuit-evasion is a well studied topic in robotics, but the results are mostly set in 2D environments, using unlimited line of sight sensing. We propose an algorithm for range limited sensing in 3D environments that captures a finite speed evader based on one single previous observation of its location. The pursuers are first moved to form a maximal cage formation, based on their number and sensor ranges, containing all of the possible evader locations. The cage is then shrunk until every part of that volume is sensed, thereby capturing the evader. The pursuers need only limited sensing range and low bandwidth communication, making the algorithm well suited for an underwater environment.