Automated negotiation is an important applying field of agent theory and technology. For the current agent theoretical models have some troubles in explaining the agent's negotiation behaviors, this paper defines utility as costs and incomes coming from the transformation of the possible negotiation states. This lead to a semantic model fitting the agent's automated negotiation. Then, on the basis of the classical Belief-Desire-Intention model, a logic named BDI-U is completely proposed, which can explain the principles of the negotiating agent's reasoning process. Therefore, the model can support further design and development of negotiating agent. The work performs five steps: designing formal language, designing semantic model, explaining semantics, proposing and explaining system axioms, and the axioms' validity proof.