Pollen Robotics has officially launched Grabette, an open-source handheld system designed to streamline the collection of robot manipulation data. The release directly addresses a critical bottleneck in contemporary robotics: despite advances in policy architectures and computational power, the industry lacks access to large-scale, diverse real-world manipulation datasets. Traditional teleoperation methods require specialized hardware, laboratories, and extensive human effort, severely limiting scalability. Grabette eliminates these barriers by enabling users to capture manipulation demonstrations using only a handheld device, a camera, and standard sensors. The Grabette system functions as a fully instrumented gripper that records synchronized visual and kinematic data during manual task demonstrations. It employs a dual-camera configuration: a wide fisheye lens provides context-rich, wrist-mounted perspectives for policy training, while an RGB-D depth camera handles robust six-degree-of-freedom trajectory reconstruction using simultaneous localization and mapping algorithms. Additional magnetic encoders capture gripper states, with all inputs synchronized to a unified clock. The hardware stack relies on accessible, off-the-shelf components, including a Raspberry Pi, standard Pi cameras, and OAK-D depth sensors, ensuring that the device can be assembled and replicated at low cost. Upon completing a demonstration, users initiate a fully browser-based processing pipeline that automatically reconstructs trajectories, cleans the data, and exports it in the standardized LeRobot format. The processed datasets are then hosted on the Hugging Face Hub, facilitating seamless integration with modern machine learning frameworks. While Grabette captures the data, its robotic counterpart, the Gripette end-effector, is required to execute the trained policies. Both devices share identical hardware specifications, guaranteeing direct simulation-to-real transferability and maintaining robot-agnostic compatibility across different manipulation tasks. The initiative positions data collection as a community-driven effort rather than an isolated laboratory exercise. By lowering the barrier between task conception and dataset generation, Pollen Robotics aims to seed a collaborative, open manipulation dataset that no single institution could develop independently. The project aligns with the architectural principles established by Stanford’s Universal Manipulation Interface, prioritizing transparency, interoperability, and rapid iteration. Future developments will include Casquette, a head-mounted point-of-view capture device designed to expand egocentric data collection capabilities. The release of Grabette marks a strategic shift toward democratizing robot learning infrastructure. By decoupling data acquisition from expensive robotic rigs and proprietary pipelines, the system enables researchers, engineers, and hobbyists to contribute to a shared data commons. As the open-source repository and processing dashboard become available to the public, the project invites global participation to accelerate the training of vision-language-action models and other manipulation policies. The long-term impact hinges on sustained community engagement, with each contributed demonstration directly expanding the diversity and utility of the collective dataset.