To model the dynamics of biological and mental processes of realistic agents as dynamical systems, the structure of the physical or physiological makeup of the agent is an important factor. This paper provides a conceptual and formal analysis based on multilevel temporal factorisation where the structure, dynamics, and adaptivity of these agent processes are distinguished conceptually in a transparent way. Moreover, the multilevel temporal factorisation analysis shows the interplay of these three aspects and how that can be modeled as an adaptive dynamical system represented in a canonical network format. In this way, an agent of any order of adaptivity can be modeled according to a tower of control levels where each level models control over the level below. This is illustrated by different case studies for higher-order adaptive agent processes. One of these case studies concerns a fifth-order adaptive agent model that illustrates how due to bad environmental influences, epigenetic effects on gene expression can lead to mental disorders.