ChronoLearn is a KG–LLM framework to structure and ex- plore Arabic historical narratives. It transforms unstructured texts into knowledge graphs using an ETL-based NLP pipeline for entity and re- lation extraction, followed by schema-guided graph construction. The system integrates graph retrieval with LLM generation (GRAG) to pro- duce grounded, explainable narratives and support semantic querying. The approach is evaluated in heterogeneous Palestinian and Jordanian sources, including Nakba-related content, using both quantitative met- rics and comparative analysis. The results demonstrate improved factual grounding and structured reasoning, addressing limitations of text-only approaches in the processing of historical knowledge in Arabic.