Abstract: Objectives Heterogeneous marine unmanned clusters can efficiently collaborate to complete complex and large-scale marine tasks, which can be used as an important means for marine scientific research, resource development, national defense and rights protection. At present, the theory and technology system of heterogeneous marine unmanned clusters are still in the exploration stage, and the related research is facing a lot of problems that need to be solved urgently. In this paper, we propose a heterogeneous marine unmanned clusters agile collaboration technology, and carry out the experimental validation on the lake for its application. Methods Firstly, for the typical application scenarios of marine unmanned cluster cooperative operation, the task characteristics, network and optimization needs are analyzed, and a heterogeneous marine unmanned cluster agile cooperative architecture based on functional node groups is designed; secondly, according to the application needs and architectural characteristics, a heterogeneous marine unmanned cluster agile task planning method based on the self-organizing graph algorithm (SOM) is proposed; thirdly, based on the principle of localized autonomy and controllability, the following controller is developed Third, based on the principle of localized autonomous control, a multi-level hardware and software decoupled agile cooperative controller for marine unmanned clusters is developed. Results In the dynamic target detection and tracking test on the lake of heterogeneous unmanned cluster, the cross-domain heterogeneous nodes are able to effectively interact with the target information based on the architectural requirements, and the average distance error during the dynamic target tracking process of the strongly-powered node group is about 40 meters, and the average tracking deviation angle is about 7-10°, which is a better completion of the collaborative detection and tracking of the dynamic target. Conclusions The effectiveness and practicality of the proposed agile collaboration technology for heterogeneous marine unmanned clusters are verified through typical scenario application demonstration and validation tests.