Hub
Capturing Physical Experience for Embodied AI
Hub is a real-world data collection platform for training world models and embodied AI. It gathers egocentric data in commercial and residential environments, offering both ready-made datasets and bespoke collection services for frontier AI labs and robotics companies. The platform manages the full pipeline from capture through enrichment, annotation and quality assurance, with delivery in formats such as MCAP and LeRobot.
Its dataset categories include egocentric RGB-D video with calibrated stereo, metric depth, IMU, 6-DoF VIO and 3D hand pose; teleoperation robot trajectories with synchronized joint states, end-effector poses, gripper commands and multi-view camera streams; hand tracking with 21-keypoint pose, MANO parameters and 6-DoF wrist motion; and tactile data from instrumented gloves and sensorized grippers, including contact pressure, shear and slip signals. Custom collection projects are matched to commercial environments and contributors, with diversity caps enforced and frame-level QA before full delivery.
Hub serves AI labs and robotics companies seeking training data, as well as small and medium businesses that can contribute capture work. The network includes verified commercial sites across industries such as farms, warehouses, construction, restaurants, auto repair, hotels and workshops. Contributors can apply to join the network, with partner SLAs and revenue share arrangements mentioned.
12 alternatives to Hub
Ranked by how well each tool replaces Hub: shared features, audience, price and popularity.
A video research and product lab building small, grounded video understanding models that,
Covers 0 of 15 key features.
57 out of 100 match—MindsHub gives everyone the freedom to choose among leading AI models and providers—use AI
Covers 0 of 15 key features, has a free plan and is open source.
Free planOpen source57 out of 100 matchUsage-based- 57 out of 100 match$39/mo
- 56 out of 100 matchContact sales
- 56 out of 100 matchContact sales
Frontier training data, RL environments and benchmarks graded by experts
Covers 2 of 15 key features.
55 out of 100 matchContact salesData and environments for training computer-use AI
Covers 0 of 15 key features and is open source.
Open source52 out of 100 match—Provides live environments for training AI agents, supplying real-world MCP environments,
Covers 0 of 15 key features.
52 out of 100 match—Research with humans and AI, in one platform.
Covers 0 of 15 key features, has a free plan and is open source.
Free planOpen source51 out of 100 matchContact salesOn-demand human experts connected to AI agents in real time via MCP.
Covers 0 of 15 key features.
51 out of 100 match—