Building a License Approval System with Cursor AI: A Low‑Effort Full‑Stack Demo

In 2025 the author explores the use of AI programming tools, focusing on Cursor, to create a small license‑approval workflow system based on MyOA. The article is aimed at non‑technical readers who want to try building simple tools or applications with AI assistance.

Project background : The team needed a fast way to generate license files for 24/7 industrial customers. With limited front‑end resources, the author turned to an AI assistant to reduce development effort.

Project flow : A high‑level flowchart (shown in the original article) illustrates the steps from user request to license generation.

Key features include a modern responsive UI, role‑based user management with JWT authentication, license request handling (search, pagination, upload fingerprint, export to Excel), and an admin approval process that automatically creates license files.

Technology stack :

Frontend – Vue 3, TypeScript, Element Plus UI, Vite. Backend – Go, Gin framework, MySQL, JWT, compression utilities, Excel handling.

Human effort: front‑end N person‑days, back‑end N person‑days. AI effort: less than one person‑day.

Sample request payload (JSON) used to call the MyOA API:

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

Cursor overview : Cursor is a VS Code‑based AI coding environment. The recent 0.44 version adds a “YOLO mode” that lets the AI execute generated commands automatically, moving toward a “semi‑autonomous driving” experience.

Work mode :

1. In normal mode, after the AI writes code you must manually click “Run” or “Accept”. 2. In YOLO mode, the AI combines its agent capabilities to run commands without confirmation, though it may still need user guidance for complex browser interactions.

Quick start for beginners :

0) Download the latest Cursor, enable YOLO mode via Settings → Feature. 1) Open an empty folder. 2) Press command + i (mac) to open the AI panel (Chat & Composer). Choose a model and optionally enable “agent” mode. 3) Describe the desired functionality in the Composer box and submit. 4) Interact with the AI through the chat, adding requirements, screenshots, or error logs as needed. 5) If execution fails, in YOLO mode the AI will attempt to fix the code automatically; otherwise you manually add commands or screenshots to guide it.

Practical tips :

• Click “Add to Composer” on command‑line errors to feed them back to the AI. • Paste screenshots of UI errors into the Composer. • Provide HTTP JSON examples when the AI needs to call internal APIs without full documentation.

Use cases demonstrated :

1) Building MVPs or small projects (personal sites, management systems, mini‑apps). 2) Experimenting with new tech stacks (e.g., deep‑learning algorithms) in minutes instead of weeks. 3) Leveraging multimodal interaction (flowcharts, UI screenshots) to rapidly prototype tools. 4) Understanding and learning from open‑source codebases.

Thoughts on AI programmers :

• Pure beginners can create functional apps with Cursor, but they still need basic knowledge of front‑end, back‑end, and deployment concepts. • AI cannot fully replace developers, especially when integrating with internal systems that lack documentation or have strict security requirements. • Code generated by AI should not be shipped to production without thorough review; it may lack proper abstraction, reuse, or testing. • Efficiency gains are evident for small modules, but larger projects still require substantial human oversight.

Conclusion : AI‑assisted coding dramatically speeds up development and expands the capability boundary for developers. While it won’t replace programmers entirely, it can make programming as accessible as video editing or driving, turning many users into “low‑code” creators.