Modern agriculture increasingly depends on autonomous navigation to optimize resource use, address labor shortages, and operate in dynamic, unstructured environments. This study explores the deployment of ORB-SLAM3, a robust visual SLAM algorithm, for agricultural navigation using resource-constrained embedded platforms and stereo cameras. An optimized configuration was tested across diverse real-world field scenarios. Results show that embedded systems can deliver reliable navigation, with smoother performance on higher-capacity platforms and reduced stability on more constrained devices in complex environments. These findings confirm ORB-SLAM3’s potential for agricultural robotics and underscore the need for further optimization to ensure robustness in varied field conditions. This work contributes to advancing practical visual SLAM deployment, enabling more efficient and scalable robotic solutions for agriculture.