The aim of this study was to examine the multilevel factor structure of the Supervisory Working Alliance Inventory (SWAI; Efstation et al., 1990) and shorten it for routine clinical use in the Chinese cultural context. Participants were 203 (Sample 1) and 97 (Sample 2) beginning counseling trainees from a master’s level training program in China. They were given the SWAI after every supervision session, and a subset of Sample 1 trainees also completed measures of their counseling self-efficacy before every client session. With Sample 1, multilevel exploratory factory analysis (M-EFA) showed that for the original 19-item version of SWAI, two factors emerged on the within-trainee level (Rapport and Client Focus) as proposed in Efstation et al. (1990), whereas only one General Alliance factor was identifiable on the between-trainee level. After removing poorly or double-loaded items, a six-item brief version of SWAI (SWAI-Brief) was obtained. Using Sample 2, multilevel confirmatory factor analysis and multilevel bifactor models supported the factor structure found in M-EFA with Sample 1. Empirical evidence also supported the reliability of the SWAI-Brief, as well as its criterion-related concurrent and predictive validity given its significant associations with trainee’s counseling self-efficacy at the same week and in the subsequent week. Results suggested that the six-item SWAI-Brief could be a reliable and valid measure for assessing supervisory working alliance in the Chinese cultural context and may be used to continuously track clinical supervision processes.