01
Modular leader-follower arms
6-DOF leader and follower arms you build yourself. Swap joints, links, and grippers. Calibrate in minutes, not days.
HardwareOpen-source teleoperation · physical AI
Inhabit
Teleoperation hardware for the embodied-AI era.
Modular leader-follower rigs and an open software stack for collecting high-quality robot manipulation data — at a fraction of the usual cost.
The bottleneck
Robot foundation models are starved for real-world manipulation data.
Models scale with data. But unlike text and images, robot manipulation data can't be scraped — it has to be physically demonstrated. The rigs to collect it are expensive, fragmented, and locked behind proprietary stacks. So the data stays scarce, and embodied AI stays stuck.
Inhabit fixes the collection layer. Affordable, modular hardware and an open pipeline so teams can capture demonstrations at scale.
What we built
A complete capture stack — every layer open.
02
Egocentric collection rig
Head-mounted stereo plus wrist cameras capture the operator's point of view — the perspective embodied policies actually learn from.
Capture03
Synced MCAP recording
Hardware-timestamped, multi-camera streams plus joint states written to MCAP. Frame-accurate sync, ready for training.
Software04
Open-source stack
CAD, firmware, drivers, and the recording pipeline — all open. Fork it, extend it, ship your own variant.
OpenBy the numbers
Built lean. Built to scale.
~$360BOM
Per leader-follower pair
4cam
Synchronized streams
6DOF
Per arm, fully articulated
100%
Open source · MIT
Start collecting data this week.
Clone the repo, print the parts, order the BOM. Everything you need is open.