Purpose: The aim of the study was to develop the Korean version of the Working Alliance Inventory (KWAI) and to assess reproducibility of both its therapist and client forms. Methods: The 12-item WAI was translated into Korean. The translation procedure followed the guidelines proposed by Beaton et al. (2000). Development of KWAI involved its translation, back-translation, and pilot testing of the pre-final version to establish Korean cultural adaptation of the original version of WAI. For test-retest reliability, therapists (N=30) and clients (N=47) completed the questionnaire on two different occasions. The interval between occasions was two to seven days, depending on subject availability. Data were collected from clinicians working in general hospitals and private clinics, and clients who had received treatment within three months. The intra-class correlation (ICC ($_{2.1}$)) was calculated for assessment of the reproducibility of the translated questionnaire. Results: The test-retest ICC ($_{2.1}$) of the client form and the therapist form was 0.92 (95%CI:0.86-0.95) and 0.94 (95%CI:0.89-0.97), respectively. Answers to items 2, 8, and 11 showed a tendency to be omitted due to ambiguity of meaning in Korean. Idiomatic expression was employed rather than word-to-word translation to have clear meaning of those items. Conclusion: The KWAI was successfully translated and adapted for applications to Korean clients and therapists, with a satisfactory level of reliability. Therefore, it can be suggested that the KWAI is useful in providing a reliable assessment of working alliance between therapists and clients.