Transformer-Based Digital Learning Assistant with Model Context Protocol (MCP) Integration for Vocational High School Students

Abstract

The implementation of the Transformer architecture in digital learning assistants still faces a major challenge: artificial intelligence models have not been integrated with local school data sources, resulting in interactions that tend to be generic and non‑contextual. This study aims to design and develop a Transformer‑based digital learning assistant integrated with the Model Context Protocol (MCP) for the Informatics subject at the Vocational High School (SMK) level. The MCP serves as a bridge enabling the chatbot to securely access local databases—including learning modules, assignments, and student information—and produce personalized responses. The system was developed following the Waterfall method, utilizing the Laravel framework, MySQL database, and the OpenAI API for the language model. System evaluation was conducted through two approaches: (1) structural testing using the white‑box testing method and cyclomatic complexity metrics on 49 functional forms covering admin, teacher, and student actors; and (2) quality validation by three expert practitioners using a practicality assessment instrument. The white‑box testing results showed a 100% success rate—all execution paths in every form were successfully traversed without any logic errors. Meanwhile, expert validation categorized the application as “Very Practical,” indicating that it is suitable for use as a learning medium. The integration of MCP successfully realized a digital assistant capable of leveraging local data to provide relevant and adaptive interactions, thereby addressing the need for personalized learning at the vocational high school level.

Keywords - Asisten Digital, Arsitektur Transformer, Kecerdasan Buatan, Pembelajaran Adaptif, Sistem Manajemen Pembelajaran.