Modelling the noise exposure for a given population is essential in order to assess its potential environmental health issues and to guide authorities in their urban planning or transportation policies. This paper introduces a new method for evaluating noise exposure of large-scale populations based on MATSim, an open- source agent-based transportation simulator. A list of agents, i.e. a synthetic population, is generated using available census and travel surveys, and are used in the transport simulation. The description of the agents includes their socio-professional profile. The resulting vehicle movements are then used in NoiseModelling, an open-source CNOSSOS-EU implementation for noise emission and propagation, to compute the noise levels every time step, typically every 15mn. Agents are then followed in their daily schedule and a noise level is assigned to them every time step. The result is a 24h noise level time history for every agent. Opportunities for analysis are discussed, such as the detailed population exposure assessment and the investigation of environmental justice issues. A case study is presented and some use cases are explored. Finally, a similar coupling for air pollutants (CO2, NOx, ...) using the HBEFA emission database and the SIRANE dispersion model is under development that will enable multi-exposure assessments.