The interaction between tumor and endothelial cells is crucial to cancer metastasis and angiogenesis. We developed a novel microfluidic device to assess the cell-cell interaction quantitatively at the single cell resolution. This integrated chip offers 16 coculture experiments in parallel with controllable microenvironments to study interactions between cells dynamically. We applied this approach to model the tumor invasion using Hela cells and human umbilical vein endothelial cells (HUVECs) and monitored the migration of both. We observed the retreatment of HUVECs upon the approach of Hela cells during coculture, indicating that the interaction between two cells was mediated by soluble factors. This interaction was further analyzed through quantitatively processing the phase-contrast microscopic time-lapse images of each individual coculture chamber. We also confirmed this paracrine effect by varying the frequency of medium change. This microfluidic technique is highly controllable, contamination free, fully automatic, and inexpensive. This approach not only offers a unique way to quantitatively study the interaction between cells but also provides accurate spatial-temporal tunability of microenvironments for cell coculture. We believe this method, intrinsically high-throughput and quantitative, will greatly facilitate the study of cell-cell interactions and communications.