Introduction: The delayed grafting incompatibility leads to a time-consuming, expensive, and labor-intensive agricultural process for the evaluation and transfer of new varieties to the industry. Current diagnostic approaches predominantly rely on sampling from the graft interface, which is invasive and may disrupt normal plant development. Identifying metabolic indicators in scion leaves offers a non-invasive alternative with minimal interference to grafted plants. However, reports on metabolic biomarkers of graft incompatibility in scion leaves remain limited, particularly for the tomato (*Solanum lycopersicum* cv. Zhongshu No. 4) and pepper (*Capsicum annuum* cv. Hangjiao No. 1) grafting combination. This study aimed to identify early metabolic indicator candidates in scion leaves for the detection of delayed graft incompatibility using a widely targeted metabolomics approach. Methods: Tomato/tomato (TT, compatible self-grafting) and tomato/pepper (TP, delayed incompatible heterografting) combinations were established. The fourth true leaf of tomato scions was collected at 5 days after grafting (DAG) for metabolomic analysis. Metabolites were extracted and analyzed using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). Multivariate statistical analyses, including principal component analysis (PCA), orthogonal partial least squares discriminant analysis (OPLS-DA), and hierarchical cluster analysis (HCA), were performed to identify differential metabolites. Metabolic pathway enrichment was analyzed using the KEGG database and metabolite set enrichment analysis (MSEA). Differential metabolites were screened based on VIP > 1, |Log2FC| ≥ 1.0, and Student’s t-test p < 0.05 with Benjamini-Hochberg FDR correction. Exogenous validation experiments were conducted by treating TT-compatible grafts with selected metabolites to assess their effects on graft survival. Results: A total of 207 differential metabolites were identified between TT and TP scion leaves, comprising 68 primary metabolites and 139 secondary metabolites. Metabolic pathway enrichment analysis revealed eight significantly enriched pathways. Among these, 31 differential metabolites were further characterized, including 24 primary metabolites and 7 secondary metabolites. Specifically, 23 metabolites were up-regulated and 8 were down-regulated in TP. The up-regulated metabolites were predominantly saccharides, whereas the down-regulated metabolites were mainly associated with flavonoid biosynthesis. Exogenous application of four primary metabolites (glucose, sucrose, mannose, trehalose) and one secondary metabolite (rhoifolin) significantly reduced the survival rate of TT-compatible grafting, confirming their adverse effects on graft compatibility. Discussion: These findings demonstrate that scion leaves can serve as a viable, non-invasive material for identifying potential biomarker candidates of delayed graft incompatibility, offering comparable diagnostic value to graft interface sampling but with greater convenience. Metabolites identified in scion leaves exhibit significant effects on graft compatibility and may function as early indicators of incompatibility. Furthermore, pathway-level metabolic analysis provides more comprehensive diagnostic insights than single-metabolite profiling, suggesting that a systems-level approach to metabolic characterization is essential for understanding graft incompatibility mechanisms. Future studies should validate these metabolic signatures across multiple cultivars and time points to establish their predictive utility.