Huge advances in the field of agent technology have been achieved over the past few years. One of the reasons for attending to agent technology is the ever-growing complexity of information systems and the increasing difficulty to foresee and plan for all potentially arising situations. Consequently, a growing number of application areas are tending to use some but by no means all - aspects of the new technology. These areas include robotics, network management, simulation, digital libraries, and electronic commerce. The full use of the whole strength of the technology, as well as the research efforts, still remain restricted to a narrow area of Distributed Artificial Intelligence. Unfortunately, this means that some pressing issues of the practical applications remain outside the scope of agent research. One of those issues is robustness in unreliable error-prone or involuntarily interacting environments. On the other hand, database technology has for a long time learned to deal with robustness by developing a suitable technology, transactions. Agent systems should be able to gain in robustness if they can be based on database systems. In this paper, an implementation model for agents with layered architecture in a transactional database environment is proposed. The model overcomes the non-orthogonality in implementing the various agent functions together with the limitation imposed on concurrent execution of agent actions. An implementation example is provided to demonstrate our objective.