Agent-based models simulate individual cells as software objects, each with their own independent states and behavioral “rules” that codify our biological hypotheses. Generally, they also are tied to models of the chemical microenvironment to emulate the motion of oxygen, growth and signaling factors, and therapeutic compounds. Together, these can form a “virtual laboratory” to computational explore hypotheses of how cells communicate and coordinate by physical and chemical signals. We will introduce PhysiCell, an open-source agent-based modeling platform tailored to multicellular systems biology, show modeling examples from breast cancer hypoxia, cancer metabolism, cancer immunology, immune response to viral infection. We will showcase our next-generation techniques that automatically translate human-interpretable hypotheses into agent-based models on-the-fly for immediate exploration and refinement. We will discuss the implications for reproducibility, model reuse, and curation of libraries of cell hypotheses.