Transcranial magnetic stimulation (TMS) is used to investigate corticomotor neurophysiology associated with functional recovery in individuals with spinal cord injury (SCI). There is insufficient evidence about test–retest measurement properties of TMS in SCI. Therefore, we investigated test–retest agreement and reliability of TMS metrics representing corticomotor excitability, output, gain, map (representation), and inhibition in individuals with cervical SCI. We collected TMS metrics from biceps and triceps muscles because of the relevance of this proximal muscle pair to the cervical SCI population. Twelve individuals with chronic C3–C6 SCI participated in two TMS sessions separated by ≥ 2 weeks. Measurement agreement was evaluated using t tests, Bland–Altman limits of agreement and relative standard error of measurement (SEM%), while reliability was investigated using intra-class correlation coefficient (ICC) and concordance correlation coefficient (CCC). We calculated the smallest detectable change for all TMS metrics. All TMS metrics except antero-posterior map coordinates and corticomotor inhibition were in agreement upon repeated measurement though limits of agreement were generally large. Measures of corticomotor excitability, output and medio-lateral map coordinates had superior agreement (SEM% < 10). Metrics representing corticomotor excitability, output, and inhibition had good-to-excellent reliability (ICC/CCC > 0.75). The smallest detectable change for TMS metrics was generally high for a single individual, but this value reduced substantially with increase in sample size. We recommend use of corticomotor excitability and recruitment curve area owing to their superior measurement properties. A modest group size (20 or above) yields more stable measurements, which may favor use of TMS metrics in group level modulation after SCI.