TetherIA/aero-hand-open: A $314 Open-Source Tendon-Driven Robotic Hand

TetherIA/aero-hand-open: A $314 Open-Source Tendon-Driven Robotic Hand

Building capable robotic hands for manipulation research has historically meant navigating proprietary hardware with five-figure price tags or fabricating custom solutions from scratch. For labs, educators, and independent researchers, this creates a frustrating gap between academic budgets and the mechanical complexity required for meaningful dexterous manipulation work. TetherIA/aero-hand-open addresses this directly: it is a fully open-source, tendon-driven robotic hand designed for accessibility without sacrificing the mechanical sophistication that makes complex grasping possible. At 389 grams and a complete kit price of $314 USD, it represents a pragmatic alternative to proprietary systems.

What is TetherIA/aero-hand-open?

TetherIA/aero-hand-open is an open-source robotic hand developed by TetherIA Inc. for dexterous manipulation research. The project sits at the intersection of hardware engineering, embedded systems, and robotics software—providing a complete platform rather than a partial component.

The hand employs a tendon-driven architecture with 7 degrees of freedom across 5 fingers (16 joints total). This design choice prioritizes mechanical efficiency and natural motion profiles over the bulkier actuator-per-joint approaches common in industrial grippers. Tendon routing enables the motors to reside in the palm and forearm regions, reducing per-finger mass and improving dynamic response.

The project is maintained by TetherIA Inc. and has accumulated 819 GitHub stars and 115 forks as of the latest commit (June 30, 2026). The primary implementation language is Python^{↗ Bright Coding Blog}, with the hardware design files, firmware, and software SDK all distributed under a mixed licensing model. The hardware ecosystem includes STEP files for CAD modification, KiCad PCB designs, complete BOM documentation, and verified print profiles for FDM fabrication.

What distinguishes this project from simpler gripper designs is its integration depth: the hand ships with an independent Python SDK, ROS2 Humble compatibility, MuJoCo simulation models, and reinforcement learning tooling built on mujoco_playground. This coverage—from physical hardware to sim-to-real training pipelines—reflects TetherIA's intent to serve researchers who need end-to-end manipulation platforms rather than isolated mechanical components.

Key Features

Tendon-Driven Mechanical Architecture The tendon-driven design routes actuation through the palm to distal joints, achieving smooth motion with reduced distal mass. This matters for dynamic manipulation tasks where finger inertia directly impacts contact stability and control bandwidth.

Complete Fabrication Stack All structural components are optimized for FDM 3D printing—specifically validated on Bambu X1C with 0.4mm nozzle, 0.2mm layer height, and tree supports restricted to build plate only. The project provides both STEP files for engineering modification and pre-oriented STL assemblies for direct printing. PCB designs include Gerber exports, KiCad source files, and CPL data for assembly services.

Weight and Cost Targets The complete hand weighs 389 grams with a full kit cost of $314 USD. These specifications position it against proprietary alternatives that typically cost 10-50x more for comparable joint counts.

Software Integration Layers

• Python SDK: Independent library for direct hardware control
• ROS2 Humble: Native integration for robotics middleware workflows
• MuJoCo Simulation: High-fidelity physics models for policy development
• RL Tooling: Training infrastructure via mujoco_playground with deployment examples in ROS2

Modular Assembly Design Each finger module assembles independently and attaches to the palm afterward. This supports field maintenance, rapid finger replacement, and iterative mechanical modification without full disassembly.

Use Cases

Academic Manipulation Research Labs studying grasp taxonomy, in-hand manipulation, or tactile sensing benefit from a platform with human-comparable kinematics at equipment-budget prices. The 7-DoF design supports research into coordinated multi-finger strategies without the procurement complexity of industrial alternatives.

Sim-to-Real Policy Transfer The MuJoCo simulation models and mujoco_playground integration enable reinforcement learning researchers to train policies in simulation and deploy through the provided ROS2 bridge. This addresses a persistent pain point in robotics ML: the gap between simulated dynamics and physical hardware behavior.

Educational Robotics Programs The complete BOM, assembly documentation, and $314 kit cost make this viable for university courses in mechanism design, embedded systems, or robotics control. Students can fabricate, assemble, program, and modify the full stack^{↗ Bright Coding Blog}—impossible with sealed proprietary hardware.

Prosthetics and Assistive Device Prototyping The lightweight design (389g) and open CAD files provide a foundation for researchers exploring anthropomorphic end-effectors. The tendon architecture mirrors biological musculotendinous structures, potentially simplifying control strategy transfer from biomechanics literature.

Commercial Robot Integration Notably, the license explicitly permits integration of purchased units into commercial products—a deliberate departure from the non-commercial restrictions placed on the design files themselves.

Installation & Setup

Hardware Preparation

1. Print structural components using the verified profile:

   • Printer: Bambu X1C or equivalent (≥200×200mm bed)
   • Nozzle: 0.4mm
   • Layer height: 0.2mm
   • Supports: Tree support, build plate only
   • Material: PLA recommended for strength and dimensional accuracy

2. Source electronics and hardware per the BOM:

   [👉 Aero Hand Open – Bill of Materials](./hardware/Assembly/BOM.csv)
   

   The BOM specifies motors, tendons, bearings, fasteners, and electronic components with vendor links and quantities.

3. Assemble finger modules following the step-by-step guide:

   https://docs.tetheria.ai/docs/assembly
   

   Each finger routes tendons independently before palm attachment.

4. Fabricate or procure the PCB using files in:

   ./hardware/PCB/
   

   Includes Gerber, KiCad project, BOM, and CPL for assembly services.

Software Setup

Python SDK Installation Refer to the SDK documentation:

https://docs.tetheria.ai/docs/sdk

And the local README:

sdk/README.md

ROS2 Integration The package targets ROS2 Humble specifically. Setup instructions and source are in:

./ros2/

Full documentation: https://docs.tetheria.ai/docs/ros2

Simulation Environment MuJoCo models and RL training code reside in:

./sim_rl/simulation/       # Physics models
./sim_rl/mujoco_playground/  # Training infrastructure

Documentation: https://docs.tetheria.ai/docs/hand_sim

Real Code Examples

Running a Pre-Defined Motion Sequence

The SDK includes sequencing capabilities for coordinated gesture execution. After installing the SDK and configuring the serial port connection to the hand's ESP32 controller:

python run_sequence.py

This script coordinates joint movements into timed sequences—pinching, palm opening, or multi-finger gestures like a peace sign. The sequencing layer abstracts the underlying tendon tension control, allowing researchers to specify high-level behaviors without managing per-joint PWM commands directly.

The example scripts are located in:

sdk/examples/

ROS2 Teleoperation Setup

For integration into larger robot systems, the ROS2 package provides standard message interfaces. The ros2/ directory contains launch files and node implementations for direct joint control and state feedback. This enables the hand to participate in existing perception-planning-control pipelines without custom bridge code.

Simulation and RL Training

The mujoco_playground integration follows standard training workflows:

# Located in: sim_rl/mujoco_playground/
# Training policies for the Aero Hand Open in simulation
# Deploy trained policies via ROS2 bridge in ros2/ directory

The repository provides explicit deployment examples showing how to transfer trained policies from simulation to the physical hardware through the ROS2 interface—addressing the sim-to-real gap with validated physics parameters.

Note: The README contains these specific code references rather than extensive inline examples. Researchers should consult the linked documentation and source directories for complete implementations.

Advanced Usage & Best Practices

Tendon Maintenance Schedule The README explicitly warns that tendon-driven mechanisms require regular tension adjustment to maintain consistent motion. Plan for periodic re-tensioning as part of experimental protocols—neglecting this produces position-dependent compliance variations that compromise repeatability.

Print Orientation Fidelity The validated print settings are specified precisely for mechanical tolerance reasons. Deviating from the recommended orientation increases support contact on functional surfaces, degrading joint clearances and tendon channel smoothness. For research reproducibility, document any print parameter deviations.

Thermal and Load Limits The disclaimer notes that overloading joints or applying excessive torque risks mechanical deformation or servo damage. The 3D-printed construction has different thermal and fatigue characteristics than machined metal—design experiments within conservative operational envelopes, particularly for sustained grasping tasks.

License Compliance for Derivatives The mixed license requires attention: software modifications fall under Apache-2.0 (permissive), while hardware design derivatives must comply with CC BY-NC-SA 4.0 (non-commercial, share-alike). Commercial manufacturing requires explicit licensing from TetherIA.

Comparison with Alternatives

The Shadow Hand offers superior kinematic complexity but at roughly 500x the cost and 10x the weight—viable for well-funded labs but prohibitive for educational or high-unit-count research. The Allegro Hand occupies a middle ground with solid performance but closed hardware and significant cost. TetherIA/aero-hand-open trades absolute DoF count for accessibility and integrated learning tooling, making it suitable for researchers prioritizing experimental throughput over anthropomorphic fidelity.

FAQ

What 3D printer is required? Any FDM printer with a ≥200×200mm bed. PLA is recommended for strength and dimensional accuracy.

Can I use this in commercial products? Yes, if you purchase units from TetherIA. Commercial manufacturing from design files requires a separate license.

What ROS2 version is supported? ROS2 Humble specifically. Other distributions are not documented.

How do I adjust tendon tension? The assembly guide covers routing and tensioning. Plan for periodic re-adjustment as normal maintenance.

Is simulation support limited to MuJoCo? Currently yes, with additional platforms under active development per the README.

What if a printed part breaks? Modular finger design allows individual replacement. Re-print from provided STLs or modify STEP files.

Where do I get help with build issues? GitHub Discussions for mechanical/electronic help, Discord for real-time debugging, Issues for bugs only.

Conclusion

TetherIA/aero-hand-open delivers a genuinely complete manipulation platform at a price point that changes procurement calculations for robotics research. The integration of hardware designs, firmware, Python SDK, ROS2 middleware, MuJoCo simulation, and RL training infrastructure—backed by explicit commercial-use permission for purchased units—reflects thoughtful productization of what could have remained a partial open-source release.

It is best suited for: academic labs with constrained equipment budgets, ML researchers needing sim-to-real pipelines, educators requiring full-stack student projects, and prototype builders exploring tendon-driven architectures. It is less suited for applications demanding industrial reliability ratings or maximum anthropomorphic fidelity regardless of cost.

The project is actively maintained with recent commits, growing community engagement (819 stars, 115 forks), and responsive support channels. For researchers who have previously compromised on gripper complexity due to cost constraints, this represents a viable path to meaningful dexterous manipulation experiments.

Explore the repository, documentation, and shop at https://github.com/TetherIA/aero-hand-open