The use of electronic agents in the conclusion of contracts has two significant advantages. Firstly, transaction costs can be reduced, and secondly, more complex contracts, being intractable by humans, are made possible. This article gives a motivation for the use of agents in the conclusion of contracts. Then, we discuss the legal classification of these contracts: Currently, agents can only be used as tools in the generation of declarations of intention; however, a different classification may be appropriate in the future. Suggestions from literature are discussed, leading to the conclusion that agents should be treated neither as natural nor as legal persons. Yet, treating agents as pure tools becomes more difficult as they act more and more autonomously. Possible solutions are discussed in this paper.