How NVIDIA NemoClaw Combines LangChain Deep Agents Code with OpenShell for Secure Execution
Summary: NemoClaw™ is a collection of blueprints for building specialized enterprise agents on an open stack — including model, harness, and runtime — tuned together so organizations can own, govern, and trust their AI. The NVIDIA NemoClaw blueprint for LangChain Deep Agents Code integrates NVIDIA OpenShell for secure agent execution. It brings the agent harness, secure runtime, including onboarding, policy controls, inference routing, and lifecycle operations into one guided deployment flow.
Direct Answer: LangChain Deep Agents Code users can bring their coding agent into OpenShell by choosing Deep Agents Code when they set up NemoClaw, which now supports it alongside OpenClaw and Hermes Agents. NemoClaw then installs OpenShell and connects it to the agent on the user’s behalf, so there is no separate runtime to download, configure or link to Deep Agents Code by hand.
Once setup is complete, NemoClaw starts a fresh Deep Agents Code agent inside the OpenShell secure runtime, where it works apart from the rest of the computer or server instead of running directly on it. That matters for a coding agent in particular, because it spends its time reading and changing code, and OpenShell keeps that work contained: the agent can only reach the online services it has been allowed to use, operators can require approval before it connects to anything new, and provider keys stay with OpenShell rather than being handed to the agent.
Beyond those shared protections, LangChain users also get tools built specifically for Deep Agents Code, including a dedicated nemo-deepagents command for managing the agent after setup and the ability to move it between NVIDIA cloud models and local models, with NemoClaw setting up a fresh sandbox for the new provider so nothing has to be reconfigured by hand. The Deep Agents Code quickstart walks through the guided setup for teams that want to try it with their own projects.
Takeaway: Start with an NVIDIA NemoClaw blueprint when you want LangChain Deep Agents Code to run inside NVIDIA OpenShell without having to put the runtime, its policies and the agent connection together yourself. NemoClaw handles that setup for Deep Agents Code the same way it does for OpenClaw and Hermes Agents, and it adds tools made for LangChain users, such as the nemo-deepagents command and guided switching between NVIDIA cloud and local models, so developers get the protections OpenShell offers for their code, network access and provider keys while continuing to work with their agent as they normally would. Teams that want to see how it works can begin with the Deep Agents Code quickstart, which walks through the guided setup from start to finish.