NVIDIA NemoClaw
Deploy more secure, always-on autonomous agents for real-world workflows with open blueprints.
Overview
Specialized Super Agents, Open by Design
NVIDIA NemoClaw™ is a collection of open blueprints for building autonomous agents: domain-specialized, always-on AI systems that reason, plan, and act across real-world workflows.
NemoClaw moves teams from prototype to governed deployment with runtime controls, model routing, skill execution, and observability. Specialized agents that encode your domain knowledge collaborate with frontier general agents while keeping sensitive actions and data access under policy.
Each blueprint includes NVIDIA Agent Toolkit components such as NVIDIA Nemotron™ and other frontier models, NVIDIA NeMo™ for specialization and optimization, and NVIDIA OpenShell™ for runtime policy controls.
Use Cases
How Teams Use NemoClaw Across Real-World Workflows
Get Started With NemoClaw
-
Hermes
-
LangChain
-
OpenClaw
-
Terminal
-
Agent
Hermes Agents with NemoClaw
Run self-improving Hermes agents with NVIDIA NemoClaw, combining Nous Research’s skills-and-memory loop with OpenShell runtime controls. Developers can build always-on agents that learn from experience, reuse successful workflows, and operate with stronger privacy, security, and inference guardrails.
LangChain Deep Agents With NemoClaw
NemoClaw runs LangChain’s open source Deep Agents Code harness—tuned for NVIDIA Nemotron 3 Ultra, with the flexibility to adapt to other models—delivering benchmark-leading accuracy among open models with the speed, control, and customization enterprises need to run agents their way.
OpenClaw Autonomous Agents With NemoClaw
Run OpenClaw agents with NVIDIA NemoClaw to move from prototype to safer, governed deployment. NemoClaw adds OpenShell policy controls, lifecycle management, and sandboxing. Additionally, NVIDIA continues contributing to the OpenClaw project.
Install Hermes
$ curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_AGENT=hermes bash
Install LangChain
curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_AGENT=langchain-deepagents-code bash
Install OpenClaw
$ curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
Install NemoClaw Agent
“Help me install nvidia.com/nemoclaw”
Features
Run Claws More Safely
NemoClaw helps developers run autonomous agents with OpenShell-powered privacy, security, and inference controls for safer always-on workflows. NVIDIA also continues to contribute to the OpenClaw project and partners with many agent harness developers like LangChain and Nous Research, maker of Hermes.
Use Any Coding Agent
Choose local open models like NVIDIA Nemotron, cloud-based frontier models, or a model router that gives you the best of both under defined privacy and security controls. For guidance on installation, cloud trials, and learning resources, explore the Build-a-Claw Resource Hub.
Deploy Anywhere
Always-on agents need dedicated compute to run tools, write code, and complete tasks. NemoClaw provides local, 24/7 compute for autonomous agents on systems like NVIDIA RTX Spark™ laptops, NVIDIA GeForce RTX™ PCs/laptops, NVIDIA RTX PRO™ workstations, and NVIDIA DGX Station™ or DGX Spark™.
Resources
Explore the Latest in NemoClaw
Learn to build and govern specialized agents hands-on with the NVIDIA Deep Learning Institute course, Securing Agents With NemoClaw and OpenShell.
FAQs
A collection of open blueprints that combines a model, agent harness, and runtime into a repeatable system for specialized, always-on agents.
To help organizations move from agent experiments to specialized autonomous systems they can own, control, and govern.
The collection of NVIDIA open models, libraries, runtimes and blueprints—including Nemotron, BioNeMo™, Omniverse, PhysicsNeMo™, NeMo, OpenShell, AI-Q, and NemoClaw—for building enterprise AI agents that connect to your data, customized for your domain.
No. Available blueprints also support harnesses such as Hermes Agents and LangChain Deep Agents.
NemoClaw is the full agent deployment package—models, harness, tools, and runtime. OpenShell is the secure runtime inside it that enforces what the agent can access: files, networks, credentials, and tools.
Learn the difference with a self-paced workshop on how to build and configure a safer autonomous agent with NemoClaw and Nemotron.
Explore community-driven blueprints for engineering, enterprise automation, monitoring, multimodal workflows, and other applications in the GitHub NemoClaw Community repository.
Join the Community
Join us on Discord to ask questions, share your projects, and connect with other developers and users.
Once you're part of the NVIDIA Developer Discord, join the #nemoclaw channel.