In today’s digitalization era, the complexity of cyber threats is growing, requiring new approaches to information security. Traditional cybersecurity systems based on monolithic architectures often prove ineffective in IoT environments. Multi-agent systems (MAS), built on the collective interaction of intelligent agents, offer a promising approach to autonomous, adaptive, and scalable protection. This paper examines the use of intelligent agents in multi-agent systems for IoT security, analyzing architectures, key communication protocols, modern use cases, specific challenges, and promising directions for development using artificial intelligence.