Based on the fundamental principles of science of complexity, the complexity of clusters of engineering projects is analyzed in the light of integrity, dynamics, gradation, openness and autonomy. It is believed that the quality supervision of clusters of engineering projects is a typical complex composed of several essential factors, between which there exists a complicated nonlinear relationship. The behaviors and manners of quality supervision are proposed including management evaluation, tree - shape quality supervision, model leading system, independent third - party assessment with a view to strengthen and enhance the quality supervision of clusters of engineering projects.