More intelligent and flexible intrusion detection systems (IDS) are required due to the growing sophistication and frequency of cyber-attacks, especially zero-day exploits. Because they rely on static signature databases, traditional IDS techniques frequently have trouble identifying new threats in real time. This study suggests a Hybrid Intrusion Detection System (HIDS) that combines Machine Learning (ML) methods with a Multiagent System (MAS) to improve detection precision and flexibility for known and unknown threats. While machine learning algorithms enable dynamic threat modelling through ongoing learning from network traffic patterns, the multi-agent architecture guarantees modularity, distributed processing, and Realtime coordination across multiple detection layers. To distinguish between typical and abnormal behavior, supervised and unsupervised learning models are used, enhancing the system's ability to recognize attacks that haven't been seen before the framework that is hybrid.