Legged autonomous vehicles move by executing patterns of leg-joint angles known as gaits. Synthesizing gaits by hand is a complex and time-consuming task which becomes even more challenging when the vehicle operates underwater. When operating underwater any motion of the limbs applies forces to the vehicle. Underwater gaits must therefore be constructed to mitigate these unwanted forces while meeting the desired gait properties. This paper presents an automatic gait synthesis system for underwater legged vehicles. The system utilizes a simulated annealing engine coupled with a black box hydrodynamic vehicle model to synthesize the desired gait. The resulting system is used to synthesize gaits for a simulated version of the AQUA amphibious hexapod although it is general enough to be applied to other legged vehicles.