Recently, Kolmogorov-Arnold Networks (KAN) has garnered significant attention for its superior accuracy and flexibility. Capitalizing on the successful application of attention mechanisms for object re-identification tasks, we introduce a pioneering approach utilizing KAN to implement attention mechanisms. Specifically, we propose a KAN spatial-channel merge attention (KAN-SCMA) and integrate it into off-the-shelf residual network architectures, raising KAN attention networks (KAAN). The KAN-SCMA module is composed of two sub-modules: the KAN spatial-channel refine (KSCR) module and the KAN spatial-channel fusion (KSCF) module. The KSCR module leverages KAN’s low computational requirements and its capacity to handle high-dimensional data, enabling efficient refinement of input feature maps. The second sub-module, the KSCF module, utilizes KAN’s flexibility in modeling diverse functions to extract meaningful spatial and channel information, then normalizes and integrates these attentions, aligning their statistical properties to facilitate seamless fusion. Extensive experiments on three public datasets, i.e., Market-1501, MSMT17, and VeRi-776, demonstrate that our KAAN exceeds state-of-the-art object Re-ID performance.