This paper presents a study on improving the adaptability of humanoid climbing robots in vertical environments. Although the robot is intended to be used for rescue and load-bearing in mountains or other hazardous areas in the future, this paper focuses on a path planning and local planning algorithm for humanoid robot wall climbing as the initial phase of our development. The first step is to acquire a depth map to extract accurate climbing holds on the vertical wall. Secondly, we propose a global planning algorithm for the humanoid robot using data from Kinect. During climbing, the humanoid robot utilizes the local planning algorithm, based on quasi-static equilibrium, to adjust its body posture to remain in an equilibrium state. Finally, all algorithms are evaluated with a simple, practical example for a humanoid climbing robot system, and its effectiveness is demonstrated experimentally in a real environment.