In recent years, selection-based hyper-heuristics have shown success across various problem domains, significantly enhancing performance in complex optimization tasks. To further advance this field, we propose a novel cooperative hyper-heuristic (COHH) that introduces a three-level agent-based framework incorporating two cooperative schemes: one focused on soft constraint satisfaction and another leveraging heuristic diversity. The framework supports both synchronous and asynchronous search modes, with asynchronous coordination reducing agent idle time and improving search efficiency. We evaluate COHH on the ITC2007 exam timetabling benchmark and compare it against six established approaches: COHH-6, HLS, SS, MTIS, EHH, and FTA. Our method demonstrates competitive performance, achieving best-known results in several datasets and delivering robust solutions under tight computational constraints. Notably, COHH attains strong averages and best-case solutions across diverse instances, highlighting the effectiveness of cooperative mechanisms in enhancing solution quality and stability.