Abstract: Gas desorption indices of drill cuttings are important parameters for predicting outburst hazards during cross-cut coal uncovering and coal roadway excavation. To solve the problems of the existing measuring instruments for gas desorption indices of drill cuttings, such as a low level of digitalization, complicated operation, slow measurement speed and difficulty in matching the drilling schedule, a multi-channel measurement system for gas desorption indices of coal drill cuttings based on Bluetooth Low Energy (BLE) was developed. The system adopted a distributed one-master multiple-slave architecture and consisted of one intrinsically safe handheld terminal with data reading and management functions and multiple measuring instruments for gas desorption indices of coal drill cuttings. Using ESP32S3 as the main control chip and integrating pressure and weighing modules, the measuring instrument could automatically measure the gas desorption index Δh₂ or K₁ of drill cuttings, and it could independently bind to and exchange data with the handheld terminal through a unique MAC address via BLE communication. Test results showed that the power consumption of the measuring instrument was only 362.6 mW in the standby state, 706.7 mW in the measuring state and 954.6 mW in the data transmission state, and it could work continuously for more than 8 h. For coal samples of different coal types under different adsorption pressures, the relative error of Δh₂ between the measuring instrument and the traditional instrument was 1.79%-7.93%, with an average relative error of 4.34% and a standard deviation of 1.87%. The relative error of K₁ was 2.44%-7.50%, with an average relative error of 5.05% and a standard deviation of 1.65%, indicating high measurement accuracy and precision. Under the conditions of 85% humidity and ambient temperature of 15-55 ℃, the system showed good stability in parameter measurement and BLE communication, and compared with the high-precision pressure sensor, the relative error of pressure measurement data was 0.83%-3.33%, with an average relative error of 1.67%.