A single command now installs security drivers, privacy protections, and local AI models on the world’s fastest-growing open source platform. OpenClaw was released on January 25, 2026. Austrian developer Peter Steinberger says he built the first version in about an hour. Within weeks, it had become one of the fastest-growing open source repositories in GitHub’s history, an AI agent that could run locally, organize files, write code, and browse the web without funneling data through the cloud.
See also: OpenClaw AI Agent: Critical vulnerabilities allow prompt injection
This uncontrolled access model was, for enterprise IT teams, both the goal and the problem. Nvidia's response came on Monday at the annual GTC developer conference in San Jose. The company announced NemoClaw, a system that installs on OpenClaw with a single command, adding the privacy and security infrastructure that enterprises need before trusting an autonomous agent with production data.
The key component is OpenShell, a new open-source runtime that isolates agents at the process level. It enforces policy-based controls on file access, network connections, and data management so that an agent can be productive without having complete freedom. The policies are written in YAML, which means that a development team can, for example, allow a sandbox to connect to a specific AI tool in the cloud while blocking everything else on the network.
OpenShell is included in Agent Toolkit , a broader collection of open models, runtimes, and designs for building long-lasting autonomous agents.
See also: OpenClaw AI Assistant: New cybersecurity threats from autonomous AI
NemoClaw also installs Nvidia’s open Nemotron models locally on any available dedicated hardware, including GeForce RTX PCs and laptops, RTX PRO workstations, DGX Station or DGX Spark. A privacy router then allows agents to reach cloud-based edge models when needed, while keeping security drivers in place.
This combination has been designed to allow agents to develop and learn new skills without ever stepping outside the defined limits.
Huang described the arrival of OpenClaw in terms that recall his usual presentation for transformative open-source moments, such as Linux, Kubernetes, and HTML, and said that the question he will now pose to every chief executive officer is: what is your strategy for OpenClaw?
See also: OpenClaw: How an open-source AI agent can be hijacked
NemoClaw is not exclusive to models. It can run any coding agent and collaborate with models from providers such as OpenAI and Anthropic, alongside Nvidia's Nemotron family, which runs locally for those who want to avoid cloud exposure entirely.
