Internet of Things (IoT) introduces many intelligent applications that are closely attached to humans’ daily activities. This advanced technology attempts to bridge the gap between the information world and the physical world. Recent studies investigate efficient, flexible, scalable and reliable IoT systems that not only control things and devices on behalf of humans but adaptable to humans’ preferences. However, the autonomous control of the IoT in a smart home or healthcare environment subjects to many factors such as human health, time and date. For example, peoples’ needs and behaviours during workdays differ from weekends or a young person needs and behaviours differs from an elderly person. Hence, the practical setting of a smart home entails flexible management to the autonomous control of IoT systems. This paper proposes an architecture of Adjustable-Autonomous Multi-agent IoT (AAMA-IoT) system to resolve a number of the IoT management of control and application interface challenges. The AAMA-IoT is applied in an elderly smart home simulation in which autonomous agents control passive things such as a chair or door and active things such as a television or an air conditioner. The test results show that the AAMA-IoT system controls 14 things with average activities recognition accuracy of 96.97%.