Um estudo comparativo de plataformas de desenvolvimento full-stack orientadas por IA

Abstract

This work presents a comparative analysis of AI-powered development platforms, focusing on solutions that adopt the text-to-code approach for generating full-stack web applications. The increasing presence of AI in the software development lifecycle has driven the emergence of platforms that aim to transform textual descriptions into functional code with minimal human intervention. Based on this premise, four platforms were selected that automate stages such as frontend, backend, and database creation: V0, Bolt, MetaGPT X, and Lovable. The study relied on quantitative criteria such as runtime performance, integration with external backend services, ease of code export and deployment support, and qualitative criteria such as learning curve, documentation and support, limitations of the free plan, and usability and user experience. For the analysis, applications were developed based on the same set of functional and non-functional requirements, using Supabase as the common backend infrastructure. The results show that although all platforms offer some form of automation, there are significant differences in f lexibility, robustness, and usability. V0 stood out for its accessibility and speed of development; Bolt and MGX offered more advanced features but with greater complexity; while Lovable showed high automation potential through its Agent Mode, despite still being under development. It is concluded that these tools have the potential to streamline full-stack development, although challenges such as customization limitations, dependence on external infrastructure, and lack of robust documentation still restrict their applicability in more complex projects.