TetherIA

The Aero Hand Open is a low-cost, open-source, anthropomorphic robotic hand developed by TetherIA, a Silicon Valley robotics startup, for affordable dexterous manipulation research. Weighing about 389 grams and sold as a self-assembly kit for roughly $314, it uses tendon-driven actuation in which seven servos pull low-friction cables to move 16 finger joints, recreating the coordinated motion (or synergies) of a human hand at a fraction of the cost of established research dexterous hands.^[1][2] TetherIA released the full mechanical, electronic, and software design on GitHub in September 2025, positioning the Aero Hand Open as one of the cheapest capable anthropomorphic hands available for embodied AI and robotics labs, alongside peers such as the ORCA Hand and Carnegie Mellon's LEAP Hand.^[1][14]

The word "Open" in the name distinguishes this community research release from TetherIA's separate, higher-degree-of-freedom commercial hand, which the company markets to industry customers.^[5] The Aero Hand Open's software is released under a permissive license, but its physical design files carry a non-commercial license, and TetherIA also sells fully assembled units through its own shop.^[2][3]

At a glance

Scroll sideways for more →

Company: TetherIA

TetherIA (styled TetherIA.ai) is a robotics company founded in early 2025 and based in Silicon Valley, California.^[5][12] It was co-founded by Yiwei Tao (who also goes by Evan) and Xu Dong (Joe).^[5] According to company profiles and trade coverage, Tao previously led the dexterous-hand project for Tesla's Optimus humanoid, overseeing it from engineering design through to mass production, and before that worked at Apple as a lead mechanical designer for the Vision Pro headset.^[5][8] Xu Dong is reported to have headed AI planning technology at the Chinese electric-vehicle maker XPeng and to have been a core member of Waymo's foundational-model team.^[5] TetherIA describes a team of roughly 15 engineers drawn from Stanford, UC Berkeley, Cornell, and Carnegie Mellon, with prior experience at Tesla, Apple, Waymo, and XPeng.^[5]

The company's stated mission is to make high-dexterity robotic hands affordable and accessible, combining hardware with AI-driven control so that manipulation research is not gated behind five- and six-figure hardware costs.^[7][12] In September 2025 TetherIA announced the completion of a multi-million-dollar angel funding round, its first, led by Matrix Partners China with participation from MiraclePlus; the company said the money would go toward expanding the team and advancing development of its first-generation products.^[6][13]

TetherIA operates two parallel product lines. The Aero Hand Open is the open-source research hand described here. Separately, the company is developing a commercial five-fingered, tendon-driven humanoid dexterous hand with a higher degree-of-freedom count, bundled with an "AI Copilot" control suite and a simulation-to-reality (sim-to-real) toolchain, with customer delivery targeted for early 2026.^[5] TetherIA showcased both the open and commercial hands at the Conference on Robot Learning (CoRL), held in Seoul, South Korea, from September 27 to 30, 2025.^[5]

The Aero Hand Open

The Aero Hand Open is a five-fingered manipulator sized close to a human hand, measuring roughly 198 by 95 by 53.5 mm and weighing about 389 grams, which keeps it light enough to mount on the wrist of a robot arm without a large payload penalty.^[2][3] The structural parts are 3D printed from nylon, and the design deliberately relies on off-the-shelf electronic components so that a lab can reproduce the hand with a desktop printer and standard parts rather than custom machining.^[8][9]

Tendon-driven actuation

Rather than placing a motor at each joint, the Aero Hand Open uses a tendon-driven architecture: seven coreless serial-bus servos sit in the base of the hand and pull thin cables that route through low-friction pathways to the finger joints.^[1][3] This "one motor per finger" scheme lets a single actuator flex several joints along a finger at once, so a modest actuator count produces coordinated, natural-looking grasps. The thumb receives three of the seven motors, giving it three actively controlled degrees of freedom for opposition, while each of the other four fingers is driven by one motor and is underactuated.^[3][4]

Across the whole hand there are 16 joints. Seven of these are actively controlled degrees of freedom; the remaining joints move through passive coupling, for example a 1:1 linkage between the DIP and PIP joints of a finger and an adaptive coupling at the MCP (knuckle) joint that lets the finger conform to an object on contact.^[4] The servos are backdrivable, meaning an external force can push the fingers back through the cables, and the compliant cable runs absorb impact energy. This compliance is a safety and robustness feature: the hand cushions collisions with objects, people, and hard surfaces instead of rigidly resisting them.^[1][4]

Sensing and control

The hand is driven by an ESP32-S3 microcontroller and communicates with a host computer over USB, running on a 6 V DC supply that draws up to roughly 8 to 10 A under load.^[3] Its firmware exposes position, velocity, and torque control modes, plus inline tendon-force sensing that estimates the tension in each cable, which can be used for force control and simple contact detection.^[1][4] The base configuration ships without fingertip tactile sensors; TetherIA and reviewers note that such sensors can be added by the user.^[9] TetherIA reports a fingertip force of about 10 N and an open-and-close cycle rate of roughly 1 Hz.^[3]

Open-source release

TetherIA published the Aero Hand Open on GitHub under the "TetherIA" organization, in the repository "aero-hand-open."^[2] The release is intended to be genuinely reproducible rather than a marketing gesture, and it includes:

Scroll sideways for more →

The project's software is licensed under Apache 2.0, while the mechanical and electrical design files are released under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 (CC BY-NC-SA 4.0), so individuals and labs may build and modify the hand freely but commercial manufacturing of the parts requires a separate license from TetherIA.^[2] By mid-2026 the GitHub repository had drawn more than 750 stars and over 100 forks, indicating meaningful community uptake.^[2]

Buyers who do not want to source parts and print the chassis themselves can order the hand two ways: a self-assembly kit at about $314 (roughly 250 pounds), or a fully assembled, ready-to-run unit for about $720 that ships with the internal PCB, an XT30 and USB-C connector, and a 6 V, 10 A power supply.^[2][3][16] Both left- and right-hand orientations are offered.^[3]

On the software side, the hand is designed to slot into modern robot-learning workflows. It integrates with ROS 2 (the Humble distribution), supports webcam-based teleoperation for collecting demonstrations, and is a native model in the MuJoCo physics simulator for reinforcement learning and sim-to-real transfer.^[1][4] TetherIA lists NVIDIA Isaac Sim support and integration with the LeRobot dataset and policy framework as planned additions.^[4][8]

Capabilities and demos

TetherIA introduced the Aero Hand Open with a launch video (titled "Introducing Aero Hand Open") and a set of demonstrations that were picked up by robotics and technology outlets.^[7][10] The clips show the hand performing everyday manipulation tasks that stress precision, delicacy, and coordination:

• Grasping and turning a small M5 screw (a 5 mm fastener), showing fine fingertip control.^[8][10]
• Picking up an iPhone from a flat surface, distributing the load across the palm and fingertips.^[9][11]
• Opening a soda can without crushing it, which requires controlled, compliant force.^[10][11]
• Pulling the trigger of a power drill and catching a tossed tennis ball, per the project's own demo reel.^[1]

TetherIA also reports more strenuous strength and endurance results, including lifting a full five-gallon water jug (on the order of 18 to 19 kg) using the fingertip force of the tendon drive, and completing a reliability test of about one million actuation cycles.^[1] These figures come from the company's own materials and should be read as vendor claims rather than independently benchmarked results.

Positioning among low-cost research hands

The Aero Hand Open sits at the extreme low-cost end of a growing class of open-source anthropomorphic hands built for robot manipulation and embodied AI research. For years, the practical options were expensive: the Allegro Hand costs on the order of $15,000, and the Shadow Robot Dexterous Hand runs well over $100,000, which put serious manipulation hardware out of reach for many academic and hobbyist labs.^[14]

A wave of open designs has attacked that gap. Carnegie Mellon's LEAP Hand (2023) is a 16-degree-of-freedom, direct-drive hand using Dynamixel servos that costs around $2,000 in parts.^[15] ETH Zurich's ORCA Hand (2025) is a 17-degree-of-freedom, tendon-driven hand with integrated tactile sensing that can be built for a material cost of roughly 2,000 Swiss francs.^[14] Independent hardware groups such as The Robot Studio and Kyber Labs have likewise released affordable 3D-printable hands aimed at researchers and makers.

Against these peers, the Aero Hand Open's distinguishing feature is price: at about $314 for a kit it is several times cheaper than LEAP or ORCA, while still offering tendon actuation, a MuJoCo model, ROS 2 integration, and a documented, reproducible build. It trades away some capability to get there. LEAP and ORCA offer more actively controlled degrees of freedom and, in ORCA's case, built-in tactile sensors, whereas the Aero Hand Open relies on underactuation and passive coupling to keep the actuator count (and cost) low. For labs whose priority is putting a workable, human-scale hand on a robot arm cheaply and iterating quickly in simulation, that trade is the entire point.

Limitations

The Aero Hand Open's design choices come with well-understood constraints, several of which reviewers and TetherIA itself have noted:^[9][10]

• Underactuation limits independent finger control. With only seven motors and heavy passive coupling, the hand cannot pose every joint independently the way a fully actuated 16-plus-motor hand can, so some in-hand manipulation strategies are out of reach.
• Minimal built-in sensing. The base hand has no fingertip tactile sensors and relies on tendon-force estimation rather than direct contact sensing, so touch-rich tasks require user-added hardware.
• 3D-printed nylon and cable drives introduce maintenance overhead. Printed parts can vary in tolerance between machines, tendons need periodic re-tensioning, and the structure can deform under excessive torque, so the hand demands more upkeep than a sealed commercial unit.
• Modest speed and payload. At roughly 1 Hz open-close and about 10 N of fingertip force, it targets research and light manipulation rather than fast or heavy industrial work.
• Non-commercial hardware license. Although the software is Apache 2.0, the CC BY-NC-SA design files bar commercial manufacture of the parts without a separate agreement, so the "open" release is aimed at research and personal use, not resale.
• Young company. TetherIA was founded in 2025 and, as of 2026, is early stage, so long-term support, spare-parts availability, and the trajectory of its commercial hand remain to be proven.^[6]

See also

• Dexterous hand
• Humanoid robot hands
• Tendon-driven actuation
• ORCA Hand
• The Robot Studio
• Kyber Labs
• Allegro Hand
• Shadow Robot
• Embodied AI

References

1. Aero Hand Open project site, TetherIA team. https://tetheria.github.io/aero-hand-open/ ↩
2. TetherIA, "aero-hand-open" GitHub repository. https://github.com/TetherIA/aero-hand-open ↩
3. Aero Hand Open (Assembled), TetherIA shop. https://shop.tetheria.ai/products/aero-hand-open ↩
4. Aero Hand Open documentation, TetherIA. https://docs.tetheria.ai/docs/intro/ ↩
5. "Silicon Valley Startup TetherIA.ai Unveils Next-Gen AI-Powered Dexterous Hand for Robotics Market," TMTPost, September 2025. https://en.tmtpost.com/post/7701816 ↩
6. "Tetheria.ai completed a multi-million-dollar angel round of financing," 36Kr, September 22, 2025. https://eu.36kr.com/en/newsflashes/3477272028305801 ↩
7. "Affordable Open-Source Robotic Hand Revolutionizes AI Research," RoboHorizon, September 2025. https://robohorizon.com/en-us/news/2025/09/affordable-open-source-robotic-hand-revolutionizes-ai-research/ ↩
8. "TetherIA's 3D-Printed Robotic Hand Costs Just $314," RoboHorizon, March 2026. https://robohorizon.com/en-gb/news/2026/03/tetheria-3d-robotic-hand/ ↩
9. "Open-Source Aero Hand Open is One You Can Actually Build at Home," TechEBlog. https://www.techeblog.com/open-source-aero-hand-diy-robotic-hand-tetheria/ ↩
10. "Video: Robotic hand with 16 joints perfectly opens soda can like humans," Interesting Engineering, September 23, 2025. https://interestingengineering.com/innovation/video-robotic-hand-opens-soda-can ↩
11. "Aero Hand Open Open Source Robotic Hand," Robotic Gizmos. https://www.roboticgizmos.com/aero-hand/ ↩
12. "This $300 robotic hand can open a soda, hold an iPhone, and move like a human," Supercar Blondie. https://supercarblondie.com/tech/aero-hand-open-robotic-hand-tetheria/ ↩
13. TetherIA company profile, Crunchbase. https://www.crunchbase.com/organization/tetheria ↩
14. Zhao et al., "ORCA: An Open-Source, Reliable, Cost-Effective, Anthropomorphic Robotic Hand for Uninterrupted Dexterous Task Learning," arXiv:2504.04259, 2025. https://arxiv.org/abs/2504.04259 ↩
15. LEAP Hand, Carnegie Mellon University (Shaw, Agarwal, Pathak). https://leaphand.com/ ↩
16. Aero Hand Open product listing, Humanoid.guide. https://humanoid.guide/product/aero-hand-open/ ↩