Demand responsive transport services are systems that assign users’ specific transport requests to different vehicles enabled to fulfill the required service. In order to contain costs, maximize profits or get a high service level, different planning algorithms have been described in literature. In our work we propose a multi-agent architecture, which adopts a mixed planning model (an integration of centralized and negotiation-based decision-making approaches) to support the system optimization from the point of view of a single fleet operator. The architecture provides flexible access to clients, through either customized or standard Web-based agents. KEYWORDS Multi-agent architecture, demand responsive system, transport system.