This paper presents the development of planar caging manipulation. It involves a preliminary conclusion where the targets are limited to convex objects and the finger number is limited to three. Despite the popularity of form or force closure analysis, we prefer caging as it owns merits like requiring little dynamics, reducing kinematics and affording robust breaking margins to tolerate control errors. The analysis part of this paper theoretically discusses optimization procedures that best exploits the merits from caging. The practice part presents implementation details of our proposal in real work space by employing KINECT, a low-cost depth image capture produced by PrimeSense. Especially, some artifice and strategies are discussed and compared in this part to fulfill application requirements. Experimental results show the efficacy of our analysis and its promising future.