Through millennia of natural evolution, flapping wings have evolved distinct advantages for aerial locomotion compared to human-built aircrafts. The bi-directional motor-driven flapping wing micro air vehicles (FWMAVs) can independently and seamlessly adjust each wing’s stroke trajectory, featuring the potential for high maneuverability and bird-like flight. In this paper, a physics-based simulator for bi-directional motor driven FWMAVs, based on the Bullet Physics engine, is introduced. Designed as a reinforcement learning environment, this simulator is fully compatible with OpenAI Gym and supports parallelization, portability, and cross-platform functionality, making it a versatile tool in robotic simulations. Extensive experiments are conducted to characterize components and assess wing aerodynamics, and the model’s fidelity is validated through wing kinematics comparison. Several example applications are used to demonstrate the simulator’s feasibility and realityz, including the evaluation and comparison of handcrafted wings to accelerate the standardization of wing production, and the training of a reinforcement learning-based controller for both locomotion and hovering.