This study presents RLTube, an algorithm that uses reinforcement learning (RL) to compute the deposition path for thin-walled bent tubes produced by wire-arc additive manufacturing. Rigid mathematical rules are used by state-of-the-art methods and the developed Brute Force Approach (BFA) to achieve this goal. In contrast, RLTube offers greater flexibility, adaptability and efficiency. This RL-based architecture uses 2D images of bent tubes as input, eliminating the need for additional feature extraction steps. As a result, RLTube deposition paths outperform BFA in terms of the developed evaluation criteria reflecting their quality.