OpenArm 7DOF Humanoid Arm – Open-Source Hardware for Physical AI
What OpenArm delivers and why it matters
OpenArm provides a complete, human‑scale 7‑DOF humanoid arm for under $6,500, combining high backdrivability, safety‑focused compliance, and a standardized "OpenArm Cell" environment to enable reproducible research in contact‑rich physical AI tasks.
Core hardware specifications
- Degrees of freedom: 7 DOF per arm, matching human arm kinematics.
- Payload: Up to 4 kg per arm, sufficient for manipulation of everyday objects.
- Backdrivability & compliance: Designed for safe human‑robot interaction, with low friction transmission and torque sensing.
- Cost: $6,500 USD for a complete bimanual system (assembled or DIY).
- Standardized research cell: The OpenArm Cell supplies uniform lighting, background, and camera placement, allowing experiments to be replicated worldwide.
Software ecosystem
OpenArm ships with a suite of open‑source repositories covering hardware design, simulation, control, and data collection:
| Repository | Purpose | Primary language | License |
|---|---|---|---|
openarm_hardware |
CAD files (STL, STEP, Fusion 360) | MDX | CERN‑OHL‑S‑2.0 |
openarm_description |
URDF/xacro robot description for simulators | MDX | Apache‑2.0 |
openarm_can |
Low‑level CAN bus motor driver library | TypeScript | Apache‑2.0 |
openarm_ros2 |
ROS 2 integration nodes and packages | TypeScript | Apache‑2.0 |
openarm_teleop |
Teleoperation tools (unilateral & bilateral control) | TypeScript | Apache‑2.0 |
openarm_isaac_lab |
NVIDIA Isaac Lab simulation environments and training tasks | TypeScript | Apache‑2.0 |
openarm_mujoco |
MuJoCo MJCF specifications and assets | TypeScript | Apache‑2.0 |
openarm_dataset |
Data‑recording utilities, Python API, and dataset format | TypeScript | Apache‑2.0 |
dora-openarm |
Dora dataflow nodes for inference, teleop, and data collection | TypeScript | Apache‑2.0 |
All repositories are publicly available on GitHub and linked from the main project page.
Community and contribution model
OpenArm is governed by an Apache‑2.0 license for software and CERN‑OHL‑S‑2.0 for hardware, encouraging both commercial and academic use. The project maintains a Code of Conduct, a Contributing guide, and an active Discord server for real‑time discussion. As of September 2026 the repository has:
- 3.3 k stars
- 363 forks
- 26 contributors
- 150 HN points indicating strong community interest.
"We're in continuous development and actively seeking contributors, research partners, and company collaborators to shape the next generation of practical humanoid systems. Ready to join the future of open‑source robotics?"
Highlights from the Hacker News discussion
- Potential applications: Users repeatedly mentioned laundry folding as a "killer app," reflecting both humor and genuine curiosity about household manipulation tasks. (e.g., "Crosses fingers for a general purpose laundry folding bot" – @NegativeLatency).
- Tele‑operation focus: The demo video primarily showcases tele‑operated control, prompting questions about the maturity of autonomous capabilities. (e.g., "The video shows mostly tele‑operated operation" – @donpdonp).
- Healthcare impact: One commenter highlighted remote surgery as a compelling use case for the arm’s compliance and safety features. (e.g., "This would be great to take advanced healthcare like surgery to remote communities!" – @MrPapz).
- Standardization value: The OpenArm Cell was praised for enabling reproducible experiments across labs, a critical need in physical AI research. (e.g., "Research performed using OpenArm can be reproduced around the world in consistent evaluation conditions" – project description).
- Community tone: The thread contains light‑hearted remarks (e.g., secret handshakes, Bruce Lee nunchaku), underscoring the community’s enthusiasm and willingness to explore creative extensions.
How to get started
- Visit the website – https://openarm.dev/ – for an overview, media, and purchase options.
- Read the documentation – https://docs.openarm.dev/ – for hardware assembly guides, software installation, and simulation tutorials.
- Join the Discord – https://discord.gg/FsZaZ4z3We – to ask questions, find collaborators, and stay updated on releases.
- Purchase or build – Use the "Buy Now" link on the GitHub README to order an assembled unit or obtain a DIY parts list.
Future roadmap (as inferred from the repo activity)
- Autonomous control: Ongoing work to integrate perception‑driven policies, likely leveraging the Isaac Lab and MuJoCo simulation pipelines.
- Payload improvements: Current 4 kg limit may be raised in later hardware revisions as motor and gearbox designs mature.
- Expanded sensor suite: Plans for higher‑resolution force‑torque sensing and tactile arrays are hinted at in the discussion of "highly sophisticated secret handshakes."
- Modular extensions: The open‑source nature encourages third‑party add‑ons such as custom end‑effectors, additional DOFs, or integration with other robotic platforms.
Conclusion
OpenArm democratizes access to a high‑performance humanoid arm by delivering a fully open‑source hardware and software stack at a price point comparable to commercial research arms. Its emphasis on safety, compliance, and a reproducible research environment positions it as a foundational platform for advancing physical AI, tele‑operation, and remote manipulation applications.