The paper introduces Temporal Kolmogorov-Arnold Networks (TKANs), a novel neural network architecture that combines the strengths of Recurrent Neural Networks (RNNs) and Kolmogorov-Arnold Networks (KANs). TKANs are designed to enhance multi-step time series forecasting by incorporating memory management and gating mechanisms. The architecture consists of Recurring Kolmogorov-Arnold Networks (RKANs) layers, which manage short-term memory and enable the network to learn and utilize past information. The integration of an LSTM cell within the RKAN layers allows for the capture of complex nonlinearities and the maintenance of long-term memory. The paper evaluates TKANs against traditional models such as GRU and LSTM using real market data, demonstrating superior performance in multi-step forecasting, particularly for longer time horizons. The results show that TKANs achieve higher R-squared values and exhibit better stability compared to other models, making them a promising approach for time series analysis.The paper introduces Temporal Kolmogorov-Arnold Networks (TKANs), a novel neural network architecture that combines the strengths of Recurrent Neural Networks (RNNs) and Kolmogorov-Arnold Networks (KANs). TKANs are designed to enhance multi-step time series forecasting by incorporating memory management and gating mechanisms. The architecture consists of Recurring Kolmogorov-Arnold Networks (RKANs) layers, which manage short-term memory and enable the network to learn and utilize past information. The integration of an LSTM cell within the RKAN layers allows for the capture of complex nonlinearities and the maintenance of long-term memory. The paper evaluates TKANs against traditional models such as GRU and LSTM using real market data, demonstrating superior performance in multi-step forecasting, particularly for longer time horizons. The results show that TKANs achieve higher R-squared values and exhibit better stability compared to other models, making them a promising approach for time series analysis.