We present a novel approach to flight control of weight-shift aircraft by employing a cable-driven parallel robot (CDPR) integrated with adaptive force control based on reinforcement learning. Development of weight-shift aircraft control has been sparse. Despite limited but notable efforts, modeling is hindered by parameter uncertainty stemming from the system’s nonlinear dynamics. The model-free control method introduced in this work operates without relying on the knowledge of the complex dynamics inherent to weight-shift aircraft flight control. An online reinforcement learning technique known as action dependent heuristic dynamic programming (ADHDP) is applied to the problem of coordinating the tension forces across parallel cable-driven actuators. Two adaptive learning agents perform demanded weight-shift maneuvers by coordinating torque commands, without an inverse kinematics model. The online reinforcement learning control is implemented on flight controller hardware with limited computational resources and strict timing constraints, performing real-time experiments on a kinematically equivalent surrogate two-body weight-shift mechanism. After online training in the presence of sustained disturbance events, the adaptive learning agents optimally balance against competing trajectory tracking objectives. The CDPR capably reproduces standard S-turn maneuvers, coordinating simultaneous banking and pitching speed actions. The encouraging experimental results inform future integration of the weight-shift CDPRs toward automatic flight control that is unprecedented for this class of aircraft.