Control Escolar: a school management system with Spring AI and an MCP server
Control Escolar is a school management web app: students, teachers, subjects, classrooms, schedules, grades, reports, users and roles, with form login and role-based access through Spring Security.
On top of it sits an AI assistant built with the Spring AI ChatClient. It keeps one conversation per user in a JDBC chat memory, answers questions about the school rules and calendar with RAG over a SimpleVectorStore, blocks sensitive words with SafeGuardAdvisor, streams answers over SSE and returns a structured student analysis with .entity(...). Its one write tool, registrarCalificacion, reads the user’s role from the ToolContext, so the model cannot grant itself permission.
The same tool methods are also published by an MCP server at /mcp (Streamable HTTP): 16 tools, five resources, a resource template, three prompts and argument completions, ready for MCP Inspector or any other MCP client.
Stack: Java 21, Spring Boot 4.1.1, Spring AI 2.0.1, OpenAI gpt-4o-mini and text-embedding-3-small, Spring Security, Spring Data JPA, Thymeleaf, H2 (PostgreSQL optional) and Micrometer Prometheus.
The app also runs without an OpenAI key: the MCP tools keep working and the rules search falls back to keywords. Setup is in the README.