Tsinghua University and Infinigence have jointly open-sourced RLark, a cloud-native management platform designed for embodied intelligence. Leveraging self-developed embodied device runtime and task-level cross-cluster network interconnection technologies, the platform integrates cloud computing power, edge nodes, and embodied devices into a unified resource system. It organizes operations around complete tasks, enabling the reuse of resource access, deployment, and communication configurations. Actual tests show that RLark can reduce device management time from hours to approximately 5 minutes, with cross-cluster tasks starting within 10 seconds. Additionally, the platform optimizes cross-regional communication performance, enabling a closed-loop for cross-regional data collection, training, and real-device validation. The entire capabilities of RLark are fully open, lowering the barrier to entry for using embodied infrastructure. It supports scaling embodied intelligence experiments to larger numbers of devices and more complex collaborative scenarios, with continuous iterations planned and the community invited to participate in its development.