Typical mobile robots can be customized to perform a variety of different tasks by combining in different ways a set of basic control modules, or behaviors. By contrast, most current systems for manipulation are still designed for just one specific task. In this paper, we propose a hierarchical behavior-based system that can perform several vision-based manipulation tasks by using different combinations of the same set of basic behaviors. Behaviors can run concurrently, and they are arbitrated through ``if-then'' rules. We show experiments involving object tracking, grasping and placing, both with static and moving objects.