This article presents a distributed architecture for university institutional knowledge management, integrating OWL ontologies, FIPA-ACL-compliant multi-agent systems (MAS), and Semantic Web technologies (RDF, SPARQL). An ontology engineered through the METHONTOLOGY methodology structures domain knowledge, while SPARQL-Generate facilitates automated semantic extraction. A hybrid storage solution (Jena Fuseki, GraphDB, MongoDB) paired with a React.js interface incorporating SPARQL querying and natural language processing via a novel Adaptive NLP-SPARQL Translator (ANST) module ensures scalability and user accessibility. A proprietary Semantic Query Optimization Algorithm (SQOA) enhances complex SPARQL query performance, achieving a 30% reduction in latency. Empirical validation on a dataset of 20,000 documents and 500 users demonstrates system robustness (99.3% success rate), with extensible applications in education, healthcare, and smart city ecosystems. Fig-ures 1 through 5 depict the architecture, agent interactions, and performance metrics.