Multicellular simulations of tumor growth in complex 3-D tissues, where data come from high content in vitro and bioengineered experiments, have gained significant attention by the cancer modeling community in recent years. Agent-based models are often selected for these problems because they can directly model and track cells' states and their interactions with the microenvironment. We describe PhysiCell, a specific agent-based model that includes cell motion, cell cycling, and cell volume changes. The model has been performance tested on systems of 10^5 cells on desktop computers, and is expected to scale to 10^6 or more cells on single super-computer compute nodes. We plan an open source release of the software in early 2016 at PhysiCell.MathCancer.org.