Agent Construction System (ACS) is an important measure to optimize the management of Public Investment Project (PIP), and it's the integration with the international agent construction. This paper analyzes the development status and supervision problems of ACS. At present, varieties of credit loss problems appeared in the construction industry. Agent construction as the specific subject of the public investment non-profit projects, its double identity of "the agent of the owners" and "the main profitability" makes it have a greater dishonesty risk in the case of lack of supervision than other construction market players. Therefore, this paper strengthens credit supervision. This paper constructs fuzzy comprehensive evaluation model based on the entropy of AHP, and designs dynamic credit supervision mechanism of agent construction based on the results of the evaluation. On the one hand credit supervision organization of agent construction is built, on the other hand, the implementation methods of dynamic supervision mechanism is proposed. Agent Construction credit supervision system in this paper fills the blank of credit index system and credit management in ACS, and helps to improve the management level of the Agent, promotes the development of ACS.