Temporal contextual information is crucial for the robustness of object tracking. However, the tightly coupled architecture of previous one-stream trackers leads to the interference with the target spatial features when processing rich temporal information. To alleviate the above issue, we propose a one-stream tracker with decoupled spatio-temporal feature extraction, named DSTTrack. The DSTTrack enables comprehensive temporal information modeling by jointly feeding the historical frame queue and the initial template-search image pair into the backbone network; subsequently, it designs spatial feature decoder and spatial feature reconstruction modules to restore spatial features suppressed by temporal features, addressing the inherent shortcoming of previous one-stream architecture in balancing temporal and spatial features. Extensive experiments demonstrate that DSTTrack achieves competitive performance across multiple mainstream benchmark datasets, including LaSOT, GOT-10k, TNL2K, UAV123, and NFS.