With the development and widespread adoption of WiFi technology, WiFi-based localization has become one of the most popular indoor positioning solutions. Existing WiFi fingerprinting methods require the creation of a fingerprint database in advance, while Multi-Layer Perception (MLP)-based WiFi localization offers low computational complexity but suffers from limited accuracy. Recently, researchers have introduced Kolmogorov-Arnold Networks (KAN) to improve localization performance. However, KAN models typically have a large number of parameters, leading to increased computational complexity. To address this issue, this paper proposes a lightweight KAN model for WiFi indoor localization (KANLoc), which consists of only two KANLinear layers. KANLoc retains the core idea of KAN while reducing parameter complexity by converting parameter calculations into simple matrix multiplications through the use of basis functions, thereby enhancing computational efficiency. Extensive experiments on the UJIIndoorLoc and Tampere datasets demonstrate that the proposed lightweight KAN model outperforms both MLP and standard KAN models.