NemoClaw

NemoClaw is an open-source security stack for the OpenClaw agent platform, introduced by NVIDIA at GTC 2026 (March 16, 2026) and hosted on NVIDIA Build. It is designed around three core pillars: safe agent execution, access control, and data protection — making it a foundational layer for developers who need to build trustworthy, production-ready AI agents. As of March 2026, NemoClaw is in early alpha preview and is not production-ready, but is installable via a single curl command. [NVIDIA NIM Free API Guide]

• Kernel-level sandboxing — Each agent runs in its own isolated environment using Landlock, seccomp, and network namespaces, preventing the agent from reaching the host filesystem or network outside its defined permissions.
• Rust-based policy engine — Every outbound connection is intercepted and validated against declarative YAML rules, with per-binary, per-host, and per-HTTP-method egress control. Every allow/deny decision creates an audit trail.
• Privacy router — Keeps sensitive context on-device by routing requests involving internal data to local models (such as NVIDIA NIM-hosted Nemotron), while only sending anonymized or non-sensitive requests to cloud-based frontier models — a practical implementation of data residency controls for agentic workflows.
• Credential injection — API keys are kept entirely off the sandbox filesystem and injected only at inference time, preventing agent-initiated exfiltration of credentials.
• Hardware-agnostic runtime — NemoClaw's OpenShell runtime runs on AMD, Intel, and NVIDIA hardware, though local inference with Nemotron is optimized for NVIDIA GPUs. [NVIDIA NIM Free API Guide]

NemoClaw is showcased prominently on the NVIDIA Build platform as a first-class offering. It works in conjunction with several other platform capabilities:

NVIDIA Build directs developers to "Build a Claw" as the primary call-to-action for NemoClaw, indicating a self-service onboarding flow. NemoClaw can be installed with a single command:

# Install NemoClaw (early preview — alpha)
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash

# Launch a sandboxed agent instance
nemoclaw my-assistant connect

# Check sandbox status
nemoclaw my-assistant status

# Follow agent logs
nemoclaw my-assistant logs --follow

[NVIDIA NIM Free API Guide] The DGX Station playbooks explicitly include NemoClaw setup as part of a secure personal AI agent guide, making it a recommended starting point for new users.

• NemoClaw is described as a new introduction ("Introducing NemoClaw") on the NVIDIA Build Overview page, suggesting it is a recently launched product at the time of the source. [NVIDIA Build Overview]
• Jensen Huang framed NemoClaw at GTC 2026 as "the missing infrastructure layer that makes autonomous AI agents viable for regulated industries." [NVIDIA NIM Free API Guide]
• NemoClaw is in alpha / early preview as of March 16, 2026 — APIs, configuration schemas, and runtime behavior are subject to breaking changes. NVIDIA explicitly states it should not be used in production environments. [NVIDIA NIM Free API Guide]
• No versioning or pricing information is provided in the available source material.