Axol

Key Features

• •Extra-Long Reach: With 860mm arm reach, Axol outreaches comparable platforms like YAM and WidowX. This provides access to a wider workspace without repositioning, enabling larger manipulation volumes and reducing the need for complex base movements.
• •Reduced Singularities: Full 180° pitch and yaw at the wrist dramatically reduces shoulder singularities. This results in a larger effective workspace and smoother trajectories, making motion planning more reliable and enabling complex tasks.
• •Protected Wiring: Fully internally routed cables protect against damage and ensure a clean, professional appearance. This design enhances durability in production environments and reduces maintenance downtime.
• •Open-Source SDK: Includes Python SDK and CLI with bimanual IK, VR teleop, and LeRobot bindings. All open source on GitHub, allowing users to build, train, and deploy custom AI policies without vendor lock-in.
• •VR Teleop: WebXR teleoperation from any compatible headset streams hand and elbow poses over WebSocket. Built-in data collection and recording modes facilitate demonstration collection for imitation learning.
• •Dual 7-DOF Arms: Two independent 7-DOF arms enable bimanual manipulation or double the cycle time of a single arm. This flexibility supports a wide range of applications from research to industrial automation.
• •FAKRA GMSL 2.0 Passthrough: Supports wrist-mount cameras with high-speed, low-latency machine vision. Ideal for real-time perception tasks such as visual servoing and object detection.
• •Deployment Ready: Steel, aluminum, and TPU construction ensures long-term reliability. The robot is designed for production environments and can be customized with modular components.

Use Cases

• •AI manipulation research: Researchers use Axol's VR teleop to quickly demonstrate tasks and collect data for imitation learning algorithms. The open-source SDK allows rapid iteration on policies, reducing time from concept to trained model.
• •Data collection for training: Almond's data collection service provides high-quality, labeled manipulation data in LeRobot format. This scales with fleet size and is delivered ready for training, enabling efficient dataset building.
• •VR teleoperation for remote control: Using WebXR, operators can teleoperate Axol from any compatible VR headset. This is useful for remote inspection, hazardous environment tasks, or collecting manipulation demonstrations.
• •Bimanual coordination experiments: Academic labs leverage the dual 7-DOF arms to study tasks requiring two-handed manipulation, such as assembly or object transfer. The platform's kinematic design simplifies programming complex motions.
• •Custom integration for automation: Companies integrate Axol with mobile bases or lifts for flexible factory automation. The modular design and customization services allow adaptation to specific application requirements.
• •Contact-rich manipulation tasks: The robot's robust construction and high payload enable tasks like peg-in-hole or assembly. The data collection service can capture labeled data for contact-rich tasks to train robust policies.

Who is this for?

Axol is built for AI researchers, robotics engineers, and developers working on physical AI applications. It is ideal for academic labs studying bimanual manipulation, companies deploying automation in production environments, and startups building AI-driven robotic systems. The platform also serves hobbyists and makers who need an open, powerful robotic arm for experimentation. Backed by Y Combinator and assembled in San Francisco, Axol is designed for users who value high-quality hardware, open-source software, and a ready-out-of-the-box experience. Target segments include manipulation researchers, imitation learning practitioners, VR teleoperation developers, and integration engineers looking to customize hardware for specific tasks.