An agonist-antagonist (AA) muscle pair is a set of two skeletal muscles that enables the voluntary movement of the skeleton. A skeletal muscle is an active hierarchical tissue, made of thousands of muscle fibers and millions of sarcomeres, the smallest contracting unit. Muscle fibers are activated by the firing of motor neurons, which is subject to the feedback provided by sensory organs such as muscle spindles. Our goal is to develop a simulation environment for the AA muscle pair that accounts for the reflexes sent to the motor neurons by the muscle spindles. In addition to the neural coupling, paired muscles are also coupled mechanically by joints, bones and tendons. In particular, we consider the case where paired muscles are directly connected by a tendon, as it happens after an agonist-antagonist myoneural interface (AMI) amputation intervention. Our simulations aim to provide insight into the novel AMI amputation technique, and help surgeons decide on critical parameters such as the optimal length of the tendon or the initial stretch of the muscle. We have implemented models of skeletal muscles, motor neurons, muscle spindles and a tendon in OpenDiHu, an open-source software designed for scalable neuromuscular simulations. In our model of the skeletal muscle, sarcomeres and muscle fibers are represented by a 0D mesh and a 1D mesh respectively. We also have a 3D continuum mechanics mesh where the global deformation of the muscle is computed. Due to the multi-scale nature of the muscle, sub-cycling is applied between the hierarchical levels defined by each mesh. This talk shows how we use the coupling library preCICE to couple the AA muscles and the tendon. At each muscle-tendon interface, we exchange displacements and forces using implicit coupling. Besides, the muscles are coupled directly using an explicit scheme to account for the crossed sensory feedback.