Today, in most office buildings, indoor environment is regulated by HVAC systems with schedule-based rules. While prevalent, these schedule-basedcontrol strategies have often resulted in low satisfaction rates and energy waste. Researchers have applied many advanced methods in building controls tooptimize occupant comfort and energy efficiency. However, it is still challenging to continuously integrate occupants’ personalized feedback into a controlsystem that has learning ability. This study proposes a bio-sensing and multi-agent reinforcement learning (RL) control system comprised of multipleRL agents and a negotiator. The RL agents aim to optimize thermal comfort of individual occupants based on their biological responses. The objectiveof the negotiator is to maximize the thermal comfort of a group of occupants in a shared environment and minimize energy consumption. A state-of-artreinforcement learning algorithm, double deep Q-learning, is implemented to train the control agents. The proposed control system is tested with threesimulated occupants in a room modeled by EnergyPlus. The result shows that the proposed system can achieve the optimized thermal comfort after 112simulation runs and improve the group thermal satisfaction by 59%, comparing to the typical schedule-based setpoint control.