NVIDIA livestream shows NemoClaw scaled alongside OpenClaw

NVIDIA Developer used an April 28 livestream to show how NemoClaw is being scaled with OpenClaw for larger, more automated agent deployments. (youtube.com) The livestream preview said the session would cover NemoClaw’s roadmap, including scaling work, model routing changes, and deeper integrations with other frameworks. It also said NVIDIA would discuss its contributions back to the OpenClaw project. (youtube.com) An AI agent is software that can keep running, call tools, access files, and make network requests without waiting for a new prompt each time. NVIDIA says NemoClaw is the wrapper around OpenClaw that adds onboarding, lifecycle management, and a hardened runtime. (docs.nvidia.com 1) (docs.nvidia.com 2) That hardened runtime is OpenShell, which NVIDIA describes as a sandbox with filesystem, network, and process controls. The company says NemoClaw uses it to apply policy-based privacy and security guardrails before an OpenClaw agent starts running. (docs.nvidia.com) (nvidia.com) The scaling session matters because NVIDIA is positioning NemoClaw for deployments beyond a single local demo. The livestream description pointed to multi-agent automation and policy-driven orchestration, which are the pieces teams need when several agents run at once under shared rules. (youtube.com) NVIDIA’s own documentation says NemoClaw is meant to run in clouds, on premises, on RTX PCs, and on DGX Spark systems. The same docs say it can route inference to NVIDIA Endpoints, OpenAI, Anthropic, Google Gemini, local Ollama, and other compatible endpoints. (docs.nvidia.com 1) (docs.nvidia.com 2) The company has been filling in that deployment story with step-by-step material. On April 17, NVIDIA published a tutorial showing NemoClaw and OpenClaw running on DGX Spark with local model serving and Telegram connectivity. (developer.nvidia.com) NVIDIA’s docs still label NemoClaw as alpha software in early preview, available since March 16, 2026. That means the livestream was less a product launch than a roadmap update on how NVIDIA wants to take its OpenClaw security layer from single-agent setup to broader fleet operations. (github.com) (docs.nvidia.com)