Motion planning refers to the process of translating high-level specifications of tasks into low-level sequences of control inputs for a robot's actuators. Legged robots, although more flexible with respect to wheeled robots in uneven and cluttered environments, are a very challenging application domain for motion planning. Such systems may benefit from the use of a multi-modal planner that is able to switch between discrete modes corresponding to the set of contact points of the robot with the ground. Widely available single-mode planners can then used to find a continuous trajectory through a given mode. Kinematic constraints usually require such paths to be constrained to a submanifold of the configuration space, which limits the efficacy of common single-mode planners and requires them to be adapted for this environment.