One-command setup
Installs the OpenShell runtime, open models and tools required to run always-on agents — all in a single command, without orchestrating 5 separate tools.
NemoClaw · powered by NVIDIA
Always-on AI agents with enterprise governance, deployed in one command.
NemoClaw + OpenShell + Nemotron models pre-installed on a dedicated Brazilian server. Kernel-level privacy, security sandbox and compliance from day 1. Rollin Host builds the stack on demand and keeps it running for you.
What is NVIDIA NemoClaw?
NemoClaw is NVIDIA's open source reference stack that simplifies running always-on assistants built on OpenClaw — with a single install command. It wraps the NVIDIA OpenShell runtime around the agent, adding policy-based privacy and security guardrails, giving your team real control over what the agent can and cannot do with your data.
NemoClaw uses open source models — such as NVIDIA Nemotron — alongside the OpenShell runtime, a secure environment designed to run agents in isolation. Combining powerful open source models with built-in guardrails makes it simpler and more robust for organizations to put AI agents into production.
NemoClaw and OpenShell are in early preview by NVIDIA. Rollin Host tracks every release, validates it in our own environment and only promotes already-stabilized versions to customers.
Why you need NemoClaw
Built-in safeguards
Applies configurable OpenShell controls for privacy, security and policy compliance. Kernel-level sandbox: the agent only sees what you allow it to see.
AI-powered workflows
Use Nemotron and other open source models for reasoning and task automation. No external API key dependency, no per-token billing, no data leaking to third parties.
Flexible execution
Run long-running agents locally, in the cloud or on enterprise hardware — with full control over where inference happens and where data resides.
How the stack connects
NemoClaw is a bundle: three open source layers communicating to deliver an always-on agent that respects your house rules.
OpenClaw Task Automation
NVIDIA OpenShell Security & Policy Controls
Open & Frontier Models Nemotron · Llama · Claude · GPT
Agent Logic & Tools Skills · Memory · Plugins
Data Sources
APIs & Services
Enterprise Systems
Cloud & Edge
What we deliver with the server
NemoClaw is open source, but getting it production-ready takes time: hardening, OpenShell policies, SSO integration, Privacy Router, monitoring, upgrade plan. We handle all of that for you.
NemoClaw + OpenShell pre-installed
Full stack configured on the latest stabilized release. No need to track NVIDIA's early preview yourself.
Security policies applied
OpenShell configured with kernel-level sandbox (Landlock + seccomp + network namespaces), read-only filesystem rules by default, centralized audit log.
Local Nemotron models
Nemotron-4 installed and optimized for your GPU. Inference runs on the machine — no byte leaves without your permission.
Privacy Router configured
Routing rules that decide what can go to external Claude/GPT and what must stay local — defined together with your DPO.
SSO and auditing
Integration with SAML, LDAP or OIDC from your provider (Google Workspace, Microsoft Entra, Okta). Structured, exportable audit logs.
24/7 human support + SLA
Brazilian team on call. Contractual SLA of 99.9% (standard) or 99.99% (custom) with automatic credits for any breach.
How much does NemoClaw cost
Two ways to get started — pick the one that makes sense for you.
Rollin Host Server
Setup from R$ 499
NemoClaw + OpenShell + Nemotron pre-configured on NVIDIA hardware sized for you. Hosting is billed separately according to tier (RTX PRO, DGX Spark or DGX Station).
- Ready in 5 business days
- 24/7 support included
- No lock-in
- Optional SSO
Hire via WhatsApp Reply in ~30 min · Mon–Fri 9am–6pm BRT
Request a custom quote
Fill in your infra details and our SDR Lana receives the data on WhatsApp in seconds. Quote in up to 24 business hours · technical hour R$ 220/h.
Payment via Pix, bank slip or card (up to 6×). Additional technical hours outside setup scope: R$ 220/h.
Cases where NemoClaw makes a difference
Regulated data on-prem
Healthcare, legal, finance: sensitive data that cannot leave your infra. NemoClaw runs entirely local, with Privacy Router blocking leakage by default.
Compliance and auditing
LGPD, ISO 27001, SOC 2: OpenShell generates a structured audit log of everything the agent did, with timestamp, user, payload — exportable to your SIEM.
Large engineering teams
Engineering with 50+ devs needs a strong sandbox: each agent isolated in its own namespace, with no risk of one workspace contaminating another.
High-volume always-on workloads
Agents processing thousands of messages/day (support, sales, surveillance) need dedicated hardware — NemoClaw on DGX delivers predictable latency.
Isolated multi-tenant
SaaS companies offering agents to their own customers: OpenShell isolates each tenant at kernel level — no shared memory between workspaces.
Air-gapped / offline
Operations on closed networks (defense, Industry 4.0, plants): NemoClaw runs 100% offline with local Nemotron. Zero internet dependency.
DIY install vs. NemoClaw managed by Rollin Host
Time to production 3–5 business days
Hardware Sized RTX/DGX
OpenShell policies Enterprise hardening ready
NVIDIA updates Validated before applying
SLA 99.9% or 99.99% contractual
Support Brazilian team 24/7
Request a quote Time to production 2 to 8 weeks for a solid environment
Hardware You buy, assemble and manage
OpenShell policies Learn and configure from scratch
NVIDIA updates Track early preview manually
SLA No SLA
Support GitHub issues + NVIDIA Discord
Want to start lighter?
Open Claw is the self-serve path.
If your case is personal use or a small team (up to 3 agents), managed Open Claw delivers the same engine without the OpenShell overhead — dashboard in Portuguese, 40-second install, plans from R$ 29.99/month.
See Open Claw →
Frequently asked questions about NemoClaw
What is NemoClaw?
NemoClaw is an NVIDIA open source stack that simplifies running always-on assistants built on OpenClaw — installed with a single command. It includes the OpenShell runtime, which adds policy-based privacy and security controls.
What is the difference between OpenClaw and NemoClaw?
OpenClaw is the agent platform (MIT, ideal for personal use). NemoClaw is the enterprise layer: it wraps OpenClaw with OpenShell sandbox, local Nemotron models and Privacy Router. In short: OpenClaw for small teams, NemoClaw for companies with governance requirements.
Can I install NemoClaw myself?
Yes, it is open source. But in enterprise production, configuring OpenShell, isolating workloads, integrating SSO, setting up Privacy Router and keeping the stack updated takes 4–8 weeks of engineering. Rollin Host delivers everything ready and maintains it for you.
Is NemoClaw compliant?
Yes. Nemotron models run locally, Privacy Router controls what can be sent to external clouds, and OpenShell adds auditing + sandbox isolation. Rollin Host complements this with a server in São Paulo, a DPA agreement and compliance-ready configuration.
What hardware do you use?
RTX PRO workstations (up to 96 GB VRAM), DGX Spark or DGX Station — always sized on a case-by-case basis based on concurrent agents and contracted models.
How much does it cost?
Custom project — price varies by hardware, number of agents, models and SLA. Request a quote and our team will have a proposal ready within 24 business hours.
How long does delivery take?
Standard provisioning in 3 to 5 business days after approval: hardware setup, NemoClaw + OpenShell installation, policy configuration, hardening, SSO integration and a 2-hour technical onboarding with your team.
Is NemoClaw in production or still in preview?
NemoClaw and OpenShell are in early preview by NVIDIA — they receive frequent updates and some APIs are still stabilizing. Rollin Host only promotes already-stabilized versions to customers, with a no-downtime upgrade plan.
Pronto pra hospedar seu projeto de IA?
Comece em 5 minutos. Migração gratuita, suporte 24/7 em português e garantia de reembolso em 7 dias.
