Building a License Approval Flow System with Cursor AI, Vue 3, and Go

The author introduces the motivation behind creating a license‑approval flow tool for MyOA: increasing demand for license generation in 24/7 industrial scenarios and limited developer resources, prompting the use of an AI coding assistant.

Cursor, an AI‑powered extension for VS Code, is presented as a more affordable alternative to Devin, offering features such as automatic code generation, debugging, PR creation, and a new "yolo" mode that can execute commands without manual confirmation.

To get started, the author outlines the installation of the latest Cursor version, enabling yolo mode via the settings, opening an empty folder, and invoking the AI panel with command + i . Users can describe their desired functionality in the Composer area, and Cursor will generate the necessary code, handling errors by iteratively refining the implementation.

Key practical tips include adding command‑line errors to the Composer with “Add to Composer”, uploading screenshots of UI errors, and providing API request examples when interfacing with internal services that lack public documentation.

The demonstrated project implements several features: a modern responsive UI, role‑based JWT authentication, license‑application management (search, pagination, upload, export), and an approval workflow that automatically generates license files after admin approval.

The technology stack consists of Vue 3, TypeScript, Element Plus, Vite for the frontend, and Go with the Gin framework, MySQL, JWT, compression, and Excel handling for the backend. The author estimates that the AI‑assisted effort required less than one person‑day compared to the traditional N‑person‑day effort.

{
   "work_items": [{
    "category": "4EA94BBA64024F2BA647177007A321A5",
    "process_name": "Testflow",
    "process_inst_id": "Testflow-20210629-001",
    "handler":"zhangsan"
    ......
   }]
}

Further sections discuss Cursor’s usage modes, how beginners can quickly prototype applications, experiment with new tech stacks (e.g., deep‑learning libraries), and build multimodal tools by feeding diagrams or screenshots.

Practical reflections cover the limits of AI programmers: while they can accelerate small‑module development, they cannot fully replace human developers for large, production‑grade systems due to documentation gaps, security concerns, and the need for code review, debugging, and architectural decisions.

Overall, the author concludes that AI tools like Cursor significantly boost development efficiency, especially for repetitive or prototype work, but they remain complementary to skilled engineers rather than complete replacements.