3D skin equivalents have been increasingly used in the pharmaceutical and cosmetic industries, but the troublesome operation procedure and low throughput restricted their applications as in vitro safety evaluation models. Organ-on-a-chip, an emerging powerful tool in tissue/organ modeling, could be utilized to improve the function of the skin model compared with that of traditional static skin models, as well as innovate an automatic and modular way for construction or detection. In this research, we grew and differentiated human keratinocytes within a microfluidic chip to construct an integrated epidermis-on-a-chip (iEOC) system, which is specially designed to integrate multi-culture units with integrated bubble removal structures as well as trans-epithelial electrical resistance (TEER) electrodes for barrier function detection in situ. After 14 days of culture at the air-liquid interface (ALI), the constructed epidermis-on-a-chip demonstrated histological features similar to those observed in normal human epidermis: a proliferating basal layer and differentiating spinous, granular, and cornified layers, especially the TEER value reached 3 kΩ cm2 and prevented more than 99% of Cascade Blue-607 Da permeation owing to the enhanced barrier function. Further immunofluorescence analysis also indicated typical keratin expression including keratin-14, keratin-10, loricrin, involucrin, and filaggrin. With the TEER monitoring integration in the chip, it could be convenient for scale-up high-quality epidermis-on-chip fabrication and correlated investigation. Additionally, the iEOC can distinguish all the 10 known toxins and non-toxins in irritation measurement by MTT assay, which is consistent with animal testing according to the OECD. Preliminarily detection of irritation responses like inflammatory cytokines also predicted different irritation reactions. This high fidelity epidermis-on-a-chip could be a potential alternative in in vitro skin irritation evaluation. This microchip and automated microfluidic systems also pave the way for scalable testing in multidisciplinary industrial applications.