The current exoskeleton and exosuit controllers lack direct control over biological tissue parameters such as muscle force and joint torques. This paper presents a predictive framework for controlling biological Musculotendon Unit (MTU) loads using exoskeletons. It introduces novel closed-form combined human-exoskeleton ordinary differential equation (ODE) models for simulating hopping and walking motions, which are utilized in the nonlinear model Predictive Controller (NMPC) of this framework. Through simulations, the framework demonstrates its capability to maintain MTU force below predefined thresholds. Future work aims to achieve real-time control of MTU load and joint torques during dynamic activities. The proposed framework, with the integration of NMPC, holds promise for enhancing the integration of exoskeleton technology in diverse applications, offering potential improvements in assistive and rehabilitative fields.