Liver cancer (LC) is the third frequent cause of death worldwide, so early diagnosis of liver cancer patients is crucial for disease management. Herein, we applied NH2-coated polystyrene@Fe3O4 magnetic beads (PS@Fe3O4-NH2 MBs) as a matrix material in laser desorption/ionization mass spectrometry (LDI-MS). Rapid, sensitive, and selective metabolic profiling of the native biofluids was achieved without any inconvenient enrichment or purification. Then, based on the selected m/z features, LC patients were discriminated from healthy controls (HCs) by machine learning, with the high area under the curve (AUC) values for urine and serum assessments (0.962 and 0.935). Moreover, initial-diagnosed and subsequent-visited LC patients were also differentiated, which indicates potential applications of this method in early diagnosis. Furthermore, among these identified compounds by FT-ICR MS, the expression level of some metabolites changed from HCs to LCs, including 29 and 12 characteristic metabolites in human urine and serum samples, respectively. These results suggest that PS@Fe3O4-NH2 MBs-assisted LDI-MS coupled with machine learning is feasible for LC clinical diagnosis.