Multi-issue negotiation is a key problem in electronic bargaining. Multi-agent technology provides effective solution for the problem. Presented in this paper is a multistage-based framework for multi-issue negotiation under time constraint in an incomplete information setting based on rational agents. The framework describes negotiation over the price of multi-issue. In this framework, in order to reduce complexity of the multi-issue negotiation, multi-issue negotiation is decomposed into multistage negotiation, and each stage has the same size (issues number). The order and number of stages are decided exogenously, and the order of issues in each stage is decided endogenously. The framework can give optimal negotiation agenda for a given decomposition of the bargaining problem based on the same size of stages. In particular, the framework can also strengthen the learning capability of participating agents by building learning system (LS) for them. Finally, a prototype system based on the negotiation framework is implemented, and the prototype system proves that the framework is effective.