A novel two-level cooperative fruit fly optimization algorithm(TCFOA) is proposed for the multidimensional knapsack problem(MKP) with the characteristics such as strong constraints and high complexity. The TCFOA is constructed based on the generation mechanism of primary and secondary fruit flies, and the secondary fruit flies are aivided into two types for exploration and exploitation to implement the balance of exploration and exploration. A communication strategy of fruit flies and a compensation mechanism based on global cost/performance ratio are proposed. A two-level structure is used to expand the search range and improve the quality of the primary fruit flies to obtain high quality solutions. The test experiments are conducted on two sets of MKP instances, and the TCFOA is compared with the methods from literature. The results demonstrate the advantages of the TCFOA in solving the MKP.