Professional paper
https://doi.org/10.51650/ezrvs.20.1-2.1
Prototyping Web Applications Using an Agentic AI Tool
Darko Andročec
; University of Zagreb, Faculty of Organization and Informatics, Varaždin, Croatia
Abstract
This paper investigates the possibilities of prototyping web applications using the agentic AI tool Claude Code and compares them with the existing low-code approach in the Oracle Apex environment, which is used in laboratory exercises of the Information Systems Design course at the Faculty of Organization and Informatics. The use case selected was the R1 invoice ERA model with six entities, where only an image file of the ERA model was provided as input to the tool. Based on that, Claude Code autonomously generated a relational database, populated it with test data, and developed a functional web application in the Python/Flask/SQLite technology stack, including CRUD forms for all entities, PDF reports, charts, and REST services returning data in JSON format. The paper presents the prompts used, the selected technologies, additional functionalities implemented through iterative steps, and an experiment of transferring the application to Oracle Apex. The results show that the basic application prototype was developed approximately four times faster than by manual development in Oracle Apex (an estimate based on the author’s practical experience and the development time measured in the observed case of a simpler ERA model with six entities), with the option of easily extending it with functionalities that are not trivial to implement in Apex, while the transfer to Oracle Apex revealed problems during the import of the generated application SQL script.
Keywords
agentic AI; Claude Code; web application prototype; vibe coding; Oracle Apex.
Hrčak ID:
349870
URI
Publication date:
27.7.2026.
Visits: 0 *