For many applications there is the need to handle user preferences and customize agents according to the user's specific needs. It is convenient to let the user provide elaborate specification consisting of constraints, preferences and objectives. Then, let the agent system make decisions about its actions by taking into account changes in the surrounding environment as well as the user preferences that come in real-time. In this paper we describe an agent programming language where we incorporate constraints, objectives and preferences into the BDI framework. Our work especially focuses on the use of soft constraints in an agent environment where we give a quantitative dimension to this agent deliberation process by apply c-semiring based techniques to determine the preferred solution.