This paper presents a study on dynamic vanpooling systems using multi-agent modeling. Our focus is on optimizing vanpooling services for passengers with shared destinations, particularly those commuting to work or educational institutions. To tackle the complexity of the assignment problem, we introduce an Agglomerative Hierarchical Clustering-based method for selecting optimal pickup points, minimizing passenger walking distances. Additionally, we propose a branch-and-bound algorithm for efficient van allocation. To evaluate the impact of vanpooling on traffic, we conduct simulations with varying van capacities and compare them to scenarios without ridesharing and with smaller van capacities. Real-world data from the Lyon network are used for experimentation. The results demonstrate that increasing van capacity significantly reduces total travel time, distance, and the number of trips and vehicles required. These findings underscore the potential of dynamic vanpooling to mitigate congestion and enhance travel efficiency in urban areas.