Open-Source Humanoid Hand at the $100 Level Achieves Real-Time Teleoperation

The open-source Amazing Hand project enables low-cost real-time teleoperation of a humanoid robotic hand, starting at about $100.
The article introduces a real-time teleoperation demo based on the open-source Enhanced Amazing Hand project, jointly driven by LeRobot, Hugging Face, and Pollen Robotics. The hardware design is fully open-source, and users can choose servos by budget: expensive models total about $200–300, while cheap models cost about $100, far below traditional humanoid hands. The demo shows smooth grasping and finger bending, and a build tutorial is provided for replication. The next step is integrating Amazing Hand with the SO101 arm for full-body teleoperation from hand to arm. The article also notes the limitations of cheap motors in precision and durability, and that full-body integration still needs verification in stability and latency.
A New Breakthrough in Open-Source Hardware: Amazing Hand
Recently, a developer shared a real-time teleoperation demo based on the open-source project Enhanced Amazing Hand, making low-cost humanoid robotic hand control a reality. The project is jointly driven by LeRobot, Hugging Face, and Pollen Robotics, and its hardware design is fully open-source. Users can choose different grades of servo motors based on their budget—using expensive models brings the total cost to about $200–300, while choosing inexpensive models keeps it around $100.

This cost level is highly attractive to robotics researchers and enthusiasts. Traditional humanoid robotic hands are often expensive, but Amazing Hand significantly lowers the barrier to entry through open-source design and modular component selection. The developer also produced a dedicated build tutorial video that walks through the assembly process in detail, making it easy for others to replicate.
Teleoperation Demo: From Hand to Full Body
In the released demo video, the developer shows how to control the robotic hand in real time through a teleoperation device to perform actions such as grasping and finger bending. The entire system responds smoothly and moves naturally, demonstrating the potential of open-source hardware for real-time control.
More notably, the developer has already planned the next step: integrating Amazing Hand with the SO101 robotic arm to achieve "full-body" teleoperation. This means that in the future we may see a complete low-cost humanoid robot remote operation system, with unified control from fine hand movements to arm motion. For scenarios such as remote operation, work in hazardous environments, or educational demonstrations, this solution offers a highly cost-effective option.
Teleoperation refers to the technology by which an operator remotely controls a robot to perform actions through input devices. The core challenge is mapping human intent to the robotic end effector with low latency and high fidelity. Common approaches include master-slave arm mapping, data gloves, motion capture systems, or VR controllers. Evaluation metrics usually include end-to-end latency, motion reproduction accuracy, and the presence or absence of force feedback. For low-cost solutions, because high-precision encoders and force sensors are lacking, position open-loop control or vision-based pose estimation is typically used, making them more dependent on the response consistency of the servo motors.
The Synergy of the Open-Source Ecosystem
The Amazing Hand project itself is open-source, with code and design files hosted on GitHub and maintained by Pollen Robotics. In this demo, the developer combined LeRobot (Hugging Face's robot learning framework) with open-source hardware, forming a complete chain from hardware to software. This collaborative model is accelerating the democratization of robotics technology—researchers no longer need expensive proprietary equipment to conduct research on dexterous hand manipulation.
Judging from community feedback, the low-cost and easy-to-replicate characteristics have drawn considerable attention to the project. As more developers join, Amazing Hand is expected to become one of the standard platforms for humanoid robotics introduction and teaching.
LeRobot is an open-source robot learning framework initiated by Hugging Face. It provides a unified interface for data collection, imitation learning, reinforcement learning training, and policy deployment, and comes with built-in drivers and dataset formats for a variety of common robotic hardware (such as the SO100/SO101 series). Its design goal is to enable researchers to complete the closed loop of "collect demonstration data—train policy—deploy on real hardware" with an extremely low barrier to entry, so it is naturally well suited to low-cost open-source hardware and is often used as a baseline platform for teaching and reproduction experiments.
Challenges and Prospects
Although the cost advantage is obvious, inexpensive servo motors may have limitations in precision and durability, which need to be weighed in practical applications. In addition, the integration of full-body teleoperation is still in the planning stage, and its stability and latency performance remain to be verified. However, continuous iteration by the open-source community often resolves these issues quickly.
Overall, Amazing Hand's real-time teleoperation demo demonstrates the feasibility of low-cost humanoid robot hardware. For researchers, students, or makers who want to explore dexterous robotic manipulation, this is undoubtedly an open-source project worth following.
Related articles

ClawDOS: Bringing AI Coding Capabilities to DOS
ClawDOS is an AI coding agent project for DOS, exploring the possibility of bringing AI coding assistance into resource-constrained retro computing environments. This article analyzes its positioning, use cases, and community response.

Baseten Releases Bison-1 and Groundhog-1, Topping the Decision Model Leaderboard
Baseten releases open-weight decision models bison-1 and groundhog-1 built on Gemma 4; bison-1 scores 76 on the AA Decision Index v2.1 to top the leaderboard. Both support 128k context and image input—here we analyze their capability combination, benchmark significance, and selection advice for developers.

Cline CLI v3.0.75 Released: Yolo Mode Now Default for One-Shot Runs
Cline CLI v3.0.75: one-shot runs now default to yolo mode, --timeout and Ctrl+C can interrupt the first round, read_files is no longer truncated to 8k chars, plus reasoning parameter fixes and 94 new providers.