Abstract: Precast keyed dry joints in assembled bent caps are prone to cracking and local shear failure due to insufficient shear capacity. Existing theories inadequately consider the influence of prestressed reinforcement configuration. Based on the deformation compatibility principle, this paper presents a shear capacity calculation formula for keyed dry joints incorporating key geometry, prestress magnitude and reinforcement ratio. Ashear capacity prediction model considering prestress influence is developed using machine learning. The validity of these methods is verified through six shear tests on keyed dry joints. Results show that the failure modes and shear strength are significantly influenced by several factors. Increasing the base height and the depth-to-height ratio can improve the shear performance. Existing calculation formulas considerably overestimate the actual shear capacity, with the maximum error up to 83.6%. In contrast, the modified formula and KNN prediction model proposed herein achieve high accuracy and broad applicability, with the maximum error of only 7.89%.