As social entities, intelligent agents need to collaborate with others regardless of whether they are cooperative or self-interested. Compared with cooperative agents, collaborations among self-interested agents are more complex and dynamic due to the selfish features. Self-interested agents are impelled to cooperate with others by their individual goals. In an agent team composed of self-interested agents, "common" goals of agents may change to be conflict as the environment changes. Especially in open and dynamic environments, if factors such as agent goals, task requirements and resources have been changed, a selfish agent may need to modify or even relieved the collaboration relationships with its "colleagues". Otherwise the collaboration would be conflict or even harmful to its individual goal. Therefore, it is important to include rational team forming mechanisms in self-interested multi-agent systems. Without a rational team-forming mechanism, agent teams in a system may have unreasonable or outdated compositions which obstruct (agent) team members to purchase profits or cause unnecessary resource consumptions. Focusing on general self-interested multi-agent systems, this paper suggests a flexible team forming mechanism that can enable agents to select team members with reasonable terms and objects. The flexibility of the mechanism enables agents to form more rational teams that can avoid potential benefit conflicts among self-interested team members.