Open infrastructure and developer tools for embodied AI.

ForgeLab Robotics builds open-source tools and infrastructure for developing, training, deploying, and orchestrating embodied AI systems.

Our goal is to make robotics development more modular, reproducible, and accessible — from hardware interfaces and data collection to model inference, workflow orchestration, and robot skills.

A modular runtime and development framework for embodied AI applications.

Forge connects robots, sensors, policies, models, and application logic through reusable nodes and dataflows, making it easier to build and integrate robotic systems without rewriting the entire stack for every platform.

Key areas include:

• Robot & sensor interfaces
• Dataflow-based runtime
• Policy and model integration
• Teleoperation and data collection
• Training & inference pipelines
• Reusable tools and robotic skills
• Multi-hardware support

We are building reusable components around modern embodied AI workflows, including:

• Vision-Language-Action models
• Imitation learning
• Robot data collection
• Real-world inference
• Perception and vision pipelines
• Motion and control tools
• Skill orchestration
• Agent-driven robot applications

ForgeLab aims to provide a unified development experience across different robot platforms and configurations, from desktop robotic arms to humanoid and mobile robots.

Instead of coupling algorithms to a single robot SDK, we want models and applications to be reusable across hardware whenever possible.

Robotics software today is often fragmented across hardware SDKs, research repositories, model implementations, and one-off demo code.

We believe the embodied AI ecosystem needs a layer that connects these pieces together.

Model ≠ Product. Demo ≠ System.

ForgeLab focuses on the engineering layer between research models and real-world robotic applications:

Robot / Sensors
      ↓
Hardware Abstraction
      ↓
Runtime & Dataflow
      ↓
Policies / Models / Tools
      ↓
Skills
      ↓
Agents & Applications

Our long-term goal is to make building an embodied AI application feel less like integrating a collection of unrelated research projects — and more like assembling reusable capabilities.

Our repositories will gradually cover several layers of the embodied AI stack:

• Forge — embodied AI runtime and integration framework
• Robot Drivers — reusable robot and sensor interfaces
• Model Runners — standardized inference adapters for robotics models
• Data Tools — collection, conversion, inspection, and dataset utilities
• Skills & Tools — reusable robotic capabilities
• Examples — end-to-end embodied AI demos and reference applications

More projects are coming.

ForgeLab Robotics is built for:

• Robotics developers
• Embodied AI researchers
• VLA / imitation learning developers
• Robot application engineers
• Students and educators exploring embodied AI
• Anyone trying to turn robotics research into a working system

ForgeLab Robotics is still evolving.

We welcome developers interested in embodied AI infrastructure, robotics tooling, model integration, and real-world robot applications.

Explore the repositories, open an issue, submit a pull request, or build something with Forge.

ForgeLab Robotics Build once. Run on robots.