Load transportation in micro-gravity environments is a challenging task for autonomous vehicles, but a crucial one considering the ever-increasing amount of space debris and the need for on-orbit assembly. We propose centralized and decentralized Model Predictive Controllers to transport passive loads. Realistic numerical simulations illustrate the performance of the proposed controllers for different load setpoints. While centralized controllers provide better transient performance, their computational cost makes the optimization problem intractable for more than a few agents involved in the collaborative task. On the other hand, decentralized controllers achieve faster computation time and do not suffer from the curse of dimensionality, at a small cost in performance.