The NemoClaw Stack
NemoClaw brings together security runtimes, AI models, hardware platforms, and enterprise tools into one cohesive stack.
Security Runtime
The security foundation. OpenShell provides kernel-level isolation, policy enforcement, and operator approval workflows for every agent action.
OpenShell
Kernel-level sandbox runtime. Isolates AI agent processes with configurable filesystem, network, and syscall policies.
settings Installed automatically with NemoClaw
AI Models
NemoClaw supports local Nemotron models for privacy-sensitive workloads and cloud providers for general tasks. The Privacy Router decides automatically.
Tip: The Privacy Router automatically selects the right model: Nemotron 120B for sensitive data that must stay local, cloud models (OpenAI, Anthropic) for general tasks. You set the policy, NemoClaw enforces it.
Nemotron Super 120B
Nemotron 3 Super 120B MoE — NVIDIA's flagship local inference model. 120 billion parameters with mixture-of-experts architecture.
settings nemoclaw model pull nemotron-super
Nemotron Nano 4B
Nemotron 3 Nano 4B — lightweight model for edge devices and quick local inference. Fits in 8GB VRAM.
settings nemoclaw model pull nemotron-nano
OpenAI
Route non-sensitive queries to OpenAI GPT models via the Privacy Router. Full compatibility with GPT-4 and GPT-5.
settings Set OPENAI_API_KEY in nemoclaw config
Anthropic
Route queries to Anthropic Claude models. The Privacy Router ensures sensitive data never leaves your infrastructure.
settings Set ANTHROPIC_API_KEY in nemoclaw config
Hardware Platforms
From developer workstations to enterprise datacenters. NemoClaw scales across NVIDIA's full hardware lineup.
GeForce RTX
Run NemoClaw on GeForce RTX 4090/5090 for development and personal use. Ideal for prototyping blueprints.
settings Install NVIDIA drivers 550+ and CUDA toolkit
RTX PRO
Professional-grade deployment on RTX PRO workstations. Certified for enterprise agent workflows.
settings Contact NVIDIA for RTX PRO deployment guide
DGX Station
Multi-GPU NemoClaw deployment on DGX Station. Run multiple sandboxed agents in parallel with dedicated GPU isolation.
settings nemoclaw deploy --target dgx-station
DGX Spark
NVIDIA's compact AI workstation. Perfect entry point for NemoClaw with built-in GPU and optimized drivers.
settings nemoclaw deploy --target dgx-spark
Deployment
Multiple deployment options: bare metal, Docker containers, remote GPU clusters, and Telegram bridge for mobile access.
Docker
Containerized NemoClaw deployment with GPU passthrough. Run isolated agent stacks with full OpenShell security.
settings docker run --gpus all ghcr.io/nvidia/nemoclaw:latest
Remote GPU
Connect NemoClaw to remote GPU clusters for inference. Scale up without local hardware.
settings nemoclaw config set remote-gpu-endpoint <URL>
Telegram Bridge
Access your NemoClaw agent from Telegram. Send commands and receive responses from any device.
settings nemoclaw bridge telegram --token <BOT_TOKEN>
Monitoring
Real-time visibility into agent behavior. Monitor sandbox activity, network requests, and model inference through built-in tools.
OpenShell TUI
Terminal UI for real-time sandbox monitoring. Watch agent actions, network requests, and policy decisions live.
settings openshell tui
Sandbox Logs
Comprehensive audit logging for all sandbox activity. Export to SIEM systems for compliance and forensics.
settings nemoclaw logs --format json | tee audit.log
Ecosystem Partners
Leading enterprise software vendors integrating NemoClaw security into their AI agent workflows.
Adobe
Adobe integrates NemoClaw security into Creative Cloud AI assistants. Sandboxed content generation with enterprise controls.
settings Partnership — contact Adobe enterprise sales
Salesforce
Salesforce Einstein agents secured by NemoClaw. CRM data stays protected with Privacy Router and network policies.
settings Partnership — contact Salesforce AI team
SAP
SAP Joule AI assistants with NemoClaw security layer. Enterprise ERP data protection and audit compliance.
settings Partnership — contact SAP AI team
Dell
Dell PowerEdge servers optimized for NemoClaw deployment. Validated hardware configurations for enterprise scale.
settings Contact Dell for validated NemoClaw configurations
Cisco
Cisco network security integrated with NemoClaw's network policy engine. Enterprise-grade network controls for AI agents.
settings Partnership — contact Cisco security team
LangChain
LangChain agents run inside NemoClaw sandboxes. Add security to any LangChain workflow with minimal code changes.
settings pip install nemoclaw-langchain
Want to Integrate with NemoClaw?
NemoClaw is open source (Apache 2.0). Build your own integration or join the partner program.