The paper introduces ADA-Track, an end-to-end multi-camera 3D multi-object tracking framework that combines the strengths of both the tracking-by-attention (TBA) and tracking-by-detection (TBD) paradigms. TBA entangles detection and tracking queries in a single embedding, while TBD decouples them but struggles to optimize both tasks effectively. ADA-Track introduces a learnable data association module based on edge-augmented cross-attention, leveraging appearance and geometric features. This module is integrated into the decoder layers of a DETR-based 3D detector, enabling simultaneous DETR-like query-to-image cross-attention for detection and query-to-query cross-attention for data association. By alternating between detection and association tasks, the framework effectively harnesses task dependencies. The method is evaluated on the nuScenes dataset, demonstrating superior performance compared to TBA and TBD approaches. The code is available at <https://github.com/dsx0511/ADA-Track>.The paper introduces ADA-Track, an end-to-end multi-camera 3D multi-object tracking framework that combines the strengths of both the tracking-by-attention (TBA) and tracking-by-detection (TBD) paradigms. TBA entangles detection and tracking queries in a single embedding, while TBD decouples them but struggles to optimize both tasks effectively. ADA-Track introduces a learnable data association module based on edge-augmented cross-attention, leveraging appearance and geometric features. This module is integrated into the decoder layers of a DETR-based 3D detector, enabling simultaneous DETR-like query-to-image cross-attention for detection and query-to-query cross-attention for data association. By alternating between detection and association tasks, the framework effectively harnesses task dependencies. The method is evaluated on the nuScenes dataset, demonstrating superior performance compared to TBA and TBD approaches. The code is available at <https://github.com/dsx0511/ADA-Track>.