A unified operating system that enables AI agents to control physical robots through natural language commands and multi-modal perception.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install embodied-os
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install embodied-os using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Embodied-OS?
Embodied-OS serves as the critical control hub bridging the gap between advanced AI agents and physical world robotics. By providing a unified Robot Abstraction Layer (RAL), it allows developers to interact with diverse hardware—from Boston Dynamics Spot to Franka Panda—using a single, consistent API. This integration within the ecosystem of Openclaw Skills empowers users to leverage Large Language Models like Claude and GPT for complex task planning and execution in real-world environments.
The system handles the heavy lifting of multi-modal perception, integrating vision, audio, and tactile sensing to give AI agents a comprehensive understanding of their surroundings. Whether you are automating a warehouse or building a research assistant, this skill provides the high-level action primitives and safety validators necessary for reliable physical interaction.
Embodied-OS Use Cases
- Warehouse Automation: Directing mobile manipulators to pick specific items based on visual tags and transport them across aisles.
- Elderly Care Assistance: Monitoring environments and responding to verbal requests or emergency signals via service robots.
- Research Lab Automation: Executing precise multi-step chemistry or physics experiments through dual-arm robot control.
- Natural Language Control: Controlling complex robotic systems by simply talking to an AI agent like ChatGPT or Claude.
How Embodied-OS Works
- Initialization: The EmbodiedOS core is initialized using a configuration file that defines robot platforms and perception sensors.
- Hardware Connection: The system establishes a connection to the physical or simulated robot via the Robot Abstraction Layer (RAL).
- Agent Integration: An AI Agent Interface is linked to the robot, enabling natural language command processing through models like Claude-4.
- Perception & Planning: The system processes multi-modal data while the AI decomposes high-level tasks into executable action primitives.
- Safe Execution: The Action Executor carries out commands while the Safety Validator monitors workspace bounds and force limits in real-time.
Embodied-OS Setup
To begin using this entry from the Openclaw Skills library, follow these steps:
1. Install via Clawhub
clawhub install embodied-os
2. Install Python Package
pip install openclaw-embodied-os
3. Configure Environment
Set your AI provider keys in your environment or a .env file:
export ANTHROPIC_API_KEY="sk-ant-..."
export OPENAI_API_KEY="sk-..."
Embodied-OS Data Schema & Taxonomy
The skill utilizes a structured configuration and state management system to organize robotic data:
| Component | Data Type | Description |
|---|---|---|
| robot | YAML/Object | Defines platform, model, and network endpoint (e.g., UR5e). |
| perception | Metadata | Manages camera resolutions, FPS, and sensor types. |
| safety | Constraints | Stores coordinate bounds, max velocity, and force limits. |
| agent | Model Config | Specifies the LLM model and API authentication for control. |
| state | Telemetry | Tracks joint positions, velocity, and real-time sensor feedback. |
Embodied-OS Advanced Features
- Multi-Layer Safety System: Real-time collision avoidance, workspace bounding, and force limit monitoring to protect hardware.
- AI-Powered Task Decomposition: Automatically breaks down complex verbal instructions into sequential robot primitives.
- Unified Robot Abstraction Layer: Swap between physical robots and simulated environments without changing your core application code.
- Multi-Modal Sensor Fusion: Synchronized processing of vision, audio, and tactile data for high-fidelity environment awareness.
SKILL.md
Loading
Related Openclaw Skills
Due Diligence Analyst
An intelligent AI agent that automates multi-dimensional due diligence for investment decisions and risk assessment.
Academic Mentor
An AI-driven research advisor designed to assist graduate students and researchers through the entire lifecycle of academic projects.
Exec Inspector
A comprehensive command-line utility for tracking, auditing, and analyzing the execution history of OpenClaw Agent tool calls.
Stock Screener & Growth Stock Analyzer
A high-performance growth stock screening and analysis tool that transforms an AI agent into a senior equity research analyst.
Enterprise Agent OS
A robust orchestration layer for managing cross-system permissions, workflow automation, and data consistency across enterprise software stacks.
FA Advisor
A professional AI investment advisory agent that automates valuation, pitch deck generation, and investor matching for primary market financing.
Featured*
