Midscene.js is a tool that utilizes AI technology to simplify UI automation. It intuitively understands user interfaces and performs necessary actions through a multimodal large language model (LLM). Users only need to describe their interaction steps or the expected data format, and the AI handles the tasks. The significance of this technology lies in its substantial reduction of the maintenance difficulty associated with UI automation, minimizing the workload of script modifications due to interface restructuring while improving the efficiency and accuracy of automated testing. Midscene.js supports multiple integration methods such as browser plugins, Puppeteer, and Playwright, and provides visual reports and debugging tools. As an open-source project, Midscene.js operates under the MIT license, ensuring data safety and privacy.