An accurate and non-invasive diagnosis of the clinical stage is critical for effectively managing liver cirrhosis. This study aimed to identify serum metabolite biomarkers and clinical features that may reliably predict high-risk cirrhosis. This cross-sectional study recruited 94 cirrhotic patients (70 for identification cohort, 24 for validation cohort) from Minhang Hospital Affiliated with Fudan University between 2018 and 2021, who were analyzed by targeted quantitative metabolomics technique. Baseline clinical characteristics were collected, and different stage cirrhosis classification was performed according to the presence or absence of decompensated events. Potential metabolite biomarkers were screened, and a model for predicting the decompensation stage was created. Finally, the incidence of decompensated outcomes was analyzed. A total of 560 metabolites were detected in the identification cohort. Indole-3-propionic acid (IPA) was the most significantly decreased metabolic biomarker in the decompensated group (P<0.01, |log2FC| >2), having the strongest correlation with hyaluronic acid (r=-0.50, P<0.01). It also performed well for differentiating decompensated cirrhosis with an area under the curve (AUC) of 0.79(0.75 at internal validation). Another diagnostic model consisting of indole-3-propionic acid, hemoglobin, and albumin showed better predictive performance with an AUC of 0.97 (0.91 at internal validation). Also, 31 (44.29%) patients developed decompensated events at a median follow-up of 22.76±15.24 months. The cumulative incidence of decompensated events based on IPA subgroups (IPA <39.67ng/ml and ≥39.67ng/ml) showed a significant difference (P<0.01). "Indole-3-propionic acid" and a diagnostic model of hemoglobin and albumin can non-invasively identify cirrhotic populations at risk for decompensation, aiding in future management of liver cirrhosis.