FM2C: An Agentic Workflow for Generating Low-Code Configurators from Feature Models

Abstract

Software Product Line (SPL) configuration often involves large feature models, technical terminology, and dependency constraints that are difficult to expose directly to end users. Using abstraction as the main principle of Low-Code Development (LCD), this paper demonstrates FM2C, an agentic workflow for generating Web-based configurators from UVL (Universal Variability Language) feature models. Given a UVL model, FM2C derives user-facing labels, explanations, and feature groupings. FM2C has two objectives: (i) to simplify the development of configurators, and (ii) to support feature selection through low-code techniques that make variability accessible to non-technical users. The generated configurators are implemented as React/Vite applications and can include wizards, dashboards, categorized lists, terminal-like interfaces, and predefined profiles. FM2C enables rapid prototyping because developers can generate interactive configurators, inspect how constraints are represented, and use the configurators to collect early feedback from end users. In the demonstration, we show how FM2C generates configurators for different software systems.