MagicQuill is an integrated image editing system designed to help users quickly realize their creative ideas. It starts with a clean yet powerful interface that allows users to express their thoughts through simple strokes, such as inserting elements, erasing objects, changing colors, and more. These interactions are monitored in real-time by a multimodal large language model (MLLM), predicting user intentions without requiring prompt input. Finally, we employ a robust diffusion prior, meticulously trained using a dual-branch plugin module, to precisely control editing requests.