Abstract: Humanoid robots mimic human structure and functionality, enabling locomotion and dexterous manipulation. Their high technical complexity involves bipedal locomotion, human-like manipulation, and autonomous decision-making. Breakthroughs in core technologies can be transferred to industrial, service, and special-purpose robotics, propelling industry advancement through the “spillover effect” and accelerating the scaled application of technologies. However, we currently face three major challenges: high hardware costs, difficult software development, and poor hardware-software integration. Building a comprehensive open-source ecosystem can effectively tackle these challenges, where full-stack open source encompasses open-source hardware including design specifications, 3D models, electrical schematics, and control schematics targeting biomimetic functional metrics covering appearance, dimensions, joint range of motion, mobility, perception, interaction, and task performance; open-source algorithms such as embodied operating systems optimized for embodied intelligence and reinforcement learning training frameworks addressing hardware-software decoupling, accelerating algorithm and RL model deployment while supporting embodied data collection/storage and compatibility with various human-robot interaction terminals; and embodied datasets containing motion data enabling machine learning and AI training to enhance intelligence level and autonomy.