Olares OS Now Runs NVIDIA NemoClaw, Bringing Sandboxed AI Agents to Personal Hardware

Integration lets NemoClaw permissioned, self-evolving agents run natively across every Olares OS device, including Olares One and NVIDIA DGX Spark^TM, by reusing the Kubernetes substrate already shared by both projects.

Paris, France — May 14th, 2026 — The Olares team announced at the GOSIM Paris 2026 conference that NVIDIA NemoClaw runs natively on Olares OS. With the integration complete, the NemoClaw reference blueprint and any NemoClaw-built agent application can now be installed and operated seamlessly across the full range of Olares-supported hardware, including the recently shipped Olares One desktop and NVIDIA DGX Spark.

Olares OS is the open-source personal cloud operating system that turns local hardware into a privacy-respecting alternative to public cloud services. Its users, such as homelab builders, content creators, design studios, and self-hosting teams, share a clear priority: to keep data, models, and agent activity on hardware they control. Built on a lightweight Kubernetes foundation, Olares OS already provides sandboxed application delivery, GPU resource management, single sign-on, and a curated marketplace of more than 200 one-click applications. NemoClaw layers onto that foundation, adding the NVIDIA OpenShell secure runtime for autonomous agents, letting users grant agents narrow, explicit permissions over their data, network, and tools rather than trusting them broadly.

Olares One is engineered for this end-to-end local AI workflow. The desktop ships with an NVIDIA GeForce RTX 5090 Mobile GPU and 24 GB of GDDR7 memory, and Olares OS extends first-class support across NVIDIA accelerated computing, from NVIDIA GeForce RTX GPUs and NVIDIA RTX PRO workstations to NVIDIA DGX Spark, NVIDIA DGX Station, and additional NVIDIA Blackwell systems, such as NVIDIA DGX B200 and NVIDIA DGX GB200. That range lets the same Olares OS run the entire agent stack locally on a user‘s own silicon: model inference, retrieval and embeddings, tool servers, and the NemoClaw-sandboxed agents that orchestrate them, with no data, prompts, or tool calls leaving the device unless the user’s policy explicitly allows it.

Because both projects are Kubernetes-native, the Olares team was able to reuse a substantial set of components across the two stacks. The result is a smaller surface area to maintain, faster uptake of upstream NemoClaw releases, and a consistent operator experience whether an agent is running on an Olares One in a studio, a DGX Spark on a developer‘s desk, or a Grace Blackwell system in a private rack.

“Our users came to Olares because they refuse to hand their data, their models, or their agents to someone else’s cloud,” said Dr. Michael Yuan, VP of Research at Olares. “Pairing the OpenShell permissioning and sandboxing model with Olares OS, across every NVIDIA GPU we support, from GeForce RTX to Grace Blackwell, means the entire pipeline, from inference to autonomous agent, can finally run on hardware the user owns, and inside boundaries the user has set.”

NVIDIA NemoClaw is an open-source reference stack that builds on the OpenClaw project to simplify the secure deployment of autonomous assistants. It provides guided onboarding, hardened blueprints, and installs OpenShell for layered process and network isolation, and routed inference across local and cloud models so developers can ship always-on, self-evolving agents that respect user-defined privacy and security policies.

NemoClaw on Olares OS is available today. Source code, documentation, and installation instructions are available at  http://github.com/beclab/olares and http://github.com/NVIDIA/NemoClaw.

About Olares

Olares is the open-source personal cloud operating system that gives individuals, teams, and organizations a private alternative to public cloud services. Together with Olares ID and the LarePass client, Olares OS powers a growing ecosystem of self-hosted productivity, creative, and AI applications, including the Olares One desktop computer, which began shipping in April 2026 following a successful Kickstarter campaign. Learn more at https://www.olares.com/.

Source: This announcement was first released on PR Newswire.