In contemporary radiotherapy planning (RTP), a key module leaf sequencing is predominantly addressed by optimization-based approaches. In this paper, we propose a novel deep reinforcement learning (DRL) model termed as Reinforced Leaf Sequencer (RLS) in a multi-agent framework for leaf sequencing. The RLS model offers a departure from time-consuming iterative optimization steps via large-scale training and can control over movement patterns through the design of reward mechanisms.We conducted experiments on four datasets with four metrics and compared our model with a leading optimization sequencer. Our findings reveal that the proposed RLS model indicates a significantly reduced reconstruction error, and potential faster convergence when integrated in an optimization planner. Additionally, RLS also showed promising results in a full deep learning RTP pipeline. We hope this first practical leaf sequencer using multi-agent RL can foster future research on machine learning for RTP.