This study developed an integrated agent-based simulation by leveraging the computational approaches and data-driven optimization to analyze the dynamic operations and behaviors in urban rail transit systems. In the proposed simulation framework, network and infrastructure environments are generated to establish the transit network characteristics including the topologies, link connectivity, and vehicle capacities. Giving the dynamic transit schedule, passengers’ travel behaviors are then simulated considering individual route choice and activities under the constraints of network, infrastructure, and operation conditions. A benchmark case study is conducted incorporating smartcard data and operational records to present the simulated dynamic conditions of the transit systems including travel patterns, in-vehicle crowdedness, operational capacity, and on-platform queuing. Validation and scalability methods of the AAAM are proposed to verify and validate the functionality of the AAAM and the reliability of the simulated results which indicate the majority of the differences between simulated value and data records are within 3 minutes. We further developed optimization scenarios through the AAAM to calibrate the true values of the real-time operational characteristics such as peak hour train occupancy. Results indicate the precision and practicality of the integrated agent-based simulation to capture behavioral and operational dynamics at robust, reliable, and scalable levels.