Recommended Free Tools
To run NemoClaw on a Linux VPS, provision at least 4 vCPUs, 8 GB RAM and 20 GB of free disk (16 GB RAM and 40 GB disk is a safer target), install Node.js 22.19 or newer with npm 10 or newer, install and start Docker, obtain an NVIDIA API key, run the current Linux installer, and finish the interactive onboarding wizard. The Node minimum changed with NemoClaw v0.0.82, released July 14, 2026, so older guides that specify Node 20 or Node 22.16 are obsolete.
Antihost offers Linux VPS plans with full root access, unlimited traffic on a 1 Gbps port and a 14-day free trial.
What you need before installing NemoClaw
NemoClaw is installed on the VPS host, while its default runtime uses Docker to start the gateway and supporting services. The installer can configure Docker on Linux, but the VPS still needs enough memory for several components to run concurrently during setup.
As an Amazon Associate I earn from qualifying purchases.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →| Resource | Official minimum | Recommended for a smoother setup | Why it matters |
|---|---|---|---|
| CPU | 4 vCPUs | 4 or more vCPUs | The gateway, container runtime and orchestration tasks run at the same time. |
| Memory | 8 GB RAM | 16 GB RAM | Docker, k3s and the OpenShell gateway can overlap during provisioning; 8 GB systems are vulnerable to out-of-memory kills. |
| Free disk | 20 GB | 40 GB | Images, layers, logs and working data consume space as the stack is pulled and initialized. |
| Operating system | Linux VPS | Ubuntu 22.04 or 24.04 are commonly used in third-party setup guides | The official operating-system list was not included in the current documentation excerpt, so confirm support for your selected distribution before provisioning. |
| Node.js | 22.19 or newer | Current Node 22 LTS release meeting that minimum | NemoClaw v0.0.82 raised the minimum; npm must be version 10 or newer. |
| Container runtime | Docker by default | Docker with a working non-root user setup | The installer uses Docker unless you explicitly select rootless Podman. |
Use the recommended memory and disk figures if you can. A small VPS may appear to meet the minimum on paper but still fail when all setup services start together. Swap can reduce the chance of an immediate memory kill, but it makes provisioning slower and does not replace physical RAM.
Choose cloud or local inference before you install
Cloud inference
Cloud inference is the default path and works on a modern VPS without a local GPU. You need an NVIDIA API key for the default provider; keys begin with nvapi-. Create the key at build.nvidia.com and keep it available for the onboarding prompt.
Local inference
Third-party setup guides report that a GPU is required only when you choose local inference. Treat that as an implementation option rather than a universal NemoClaw requirement: a CPU-only VPS is suitable for the default cloud path, while a local model requires a GPU-enabled host with enough VRAM for the model you select. Confirm the model’s hardware requirements separately before renting a GPU instance.
Docker or rootless Podman
Docker is the default runtime and has the broadest path through the installer. Linux users who already standardize on rootless Podman can select it by setting NEMOCLAW_GATEWAY_RUNTIME=podman. Podman does not require the Docker CLI or daemon, but you must have a functioning rootless Podman installation before starting NemoClaw. Choose the runtime your team already operates unless you have a specific reason to change it.
| Decision | Choose this when | Extra consideration |
|---|---|---|
| Docker | You want the default, documented path or need the installer to configure the runtime. | Docker setup may require elevated privileges even though the NemoClaw installer itself does not require root. |
| Rootless Podman | Your VPS already uses Podman and you do not want a Docker daemon. | Set NEMOCLAW_GATEWAY_RUNTIME=podman and verify rootless networking and image access first. |
| Cloud inference | You have no GPU or want the simplest VPS deployment. | Provide an NVIDIA API key during onboarding and account for provider usage costs. |
| Local inference | You need local model execution and have a compatible GPU host. | GPU, driver, VRAM and model compatibility become your responsibility. |
Provision and harden the VPS
- Create a Linux VPS with at least 4 vCPUs, 8 GB RAM and 20 GB free disk. Prefer 16 GB RAM and 40 GB free disk for regular use.
- Use a current Ubuntu image if you want to follow the most common third-party examples: Ubuntu 22.04 and 24.04 are the versions those guides recommend. Because the official OS matrix is not reproduced here, check NVIDIA’s current prerequisites page for a definitive support statement.
- Connect over SSH using a normal administrative account. Keep a separate root or sudo-capable account for operating-system changes, but run NemoClaw’s installer as your normal user.
- Apply your distribution’s security updates, set the correct time zone, and configure the VPS provider’s firewall before exposing any service. Do not open public ports merely because the installer is running.
- Record the VPS’s available memory and disk so you can distinguish a resource failure from a package or permission failure later. On Ubuntu,
free -handdf -hprovide a quick check.
If you plan to publish a web dashboard or terminate HTTPS with Caddy later, a third-party guide recommends allowing ports 80 and 443 in the cloud firewall first. That advice is not part of the confirmed NemoClaw installation requirements; open those ports only when you have a specific service configured to use them, and restrict administrative interfaces separately.
Install the required Node.js version
The current requirement is Node.js 22.19 or later and npm 10 or later. NemoClaw v0.0.82, released July 14, 2026, introduced this minimum. Guides that say Node 20 or Node 22.16 is sufficient refer to an older requirement.
Install Node 22 with nvm
- Install or enable nvm for the user who will run NemoClaw.
- Install the Node 22 line and make it active:
nvm install 22
nvm use 22- Check both versions before running the installer:
node --version
npm --versionThe Node output must be 22.19.0 or higher, and npm must be 10.0.0 or higher. If the installer reports that your Node version is too old, install a newer Node release through nvm and run the installer again. Do not assume that a system package named nodejs is current enough.
Make the version available in your shell
nvm and fnm modify shell startup files. A new terminal may see Node while the current SSH session does not. If NemoClaw later reports a missing command, reload the appropriate file:
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →source ~/.bashrc
# or, for zsh:
source ~/.zshrcOpening a new SSH session has the same effect. Confirm which node, node --version and npm --version in the same shell from which you will run the installer.
Install and test Docker
On Linux, the NemoClaw installer can install Docker, start its service and add your user to the docker group. Docker installation itself may need sudo privileges. The NemoClaw installer does not require you to run as root.
- Install Docker using your distribution’s supported Docker instructions, or allow the NemoClaw installer to perform that setup.
- Start the Docker service if it is not already running.
- Ensure your normal user can invoke Docker without an error:
docker --version
docker psIf the setup added you to the docker group but the current shell still lacks access, run:
newgrp dockerThen retry docker ps. A permission error at this stage usually originates in Docker’s user or service configuration, not in NemoClaw. Fix Docker access first; repeatedly rerunning the NemoClaw installer will not repair a shell that cannot talk to the daemon.
Using Podman instead
Skip Docker only when you intentionally choose rootless Podman. Install and test Podman under the same user that will run NemoClaw, then export the runtime selection before installation:
export NEMOCLAW_GATEWAY_RUNTIME=podmanDo not set this variable accidentally: selecting Podman without a working rootless environment changes the failure from a Docker permission problem to a runtime-availability problem.
Get the API key for the default provider
Cloud inference requires an NVIDIA API key. Create one at build.nvidia.com; the key format begins with nvapi-. Treat it like a password: do not paste it into shell history, commit it to a repository, or include it in screenshots and support logs. The onboarding wizard asks for the provider credential after you choose the provider and model.
Run the NemoClaw installer
Use the current Linux installer command from NVIDIA’s NemoClaw documentation rather than copying a command from an older guide. The release line is changing quickly, and installer URLs and flags can change. Run the command as your normal user after Node.js and the runtime pass their checks.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsInteractive installation
For a first deployment, interactive mode is the safer choice because it lets you review the runtime, provider and policy choices. Paste the current command exactly as published, then follow the prompts. If Docker is missing, the installer can install and start it; if group access cannot be refreshed automatically, it will tell you to run newgrp docker and retry.
Scripted or automated installation
Set both variables before invoking the official installer:
export NEMOCLAW_NON_INTERACTIVE=1
export NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1The first disables prompts. The second records acceptance of the installer’s third-party software terms. Automation should still validate the Node version, runtime health, available disk and memory before launching the command, and should supply credentials through your secret-management system rather than embedding them in a script.
Do not infer a successful installation solely from a completed shell command. Continue through onboarding and perform the checks in the next sections.
Complete the onboarding wizard
Onboarding is interactive even though the underlying services are containerized. The sequence below reflects the current wizard flow.
- Select OpenClaw as the agent. This is the agent choice in the current onboarding path.
- Choose an inference provider. Select the cloud provider associated with your API key, or choose a local option only if the VPS has the required GPU and model support.
- Select a model. Pick a model that matches the provider account and the workload you intend to run.
- Enter the provider credential. Paste the NVIDIA key when using the default cloud path. Verify that it starts with
nvapi-and that you have not included surrounding spaces. - Skip optional web search and messaging on the first run. You can add those integrations after the base gateway works. Deferring them isolates installation problems from external-service configuration.
- Accept the suggested network-policy tier. The wizard proposes a policy tier; accept the suggested value for the initial deployment unless you already have a reviewed policy requirement.
If an SSH disconnect, rejected credential or resource failure interrupts the wizard, resume instead of starting over:
nemoclaw onboard --resumeThe resume command continues the onboarding state that was already created.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Verify the command and services
After onboarding, start with the CLI’s built-in help so you use the commands provided by the version actually installed on your VPS:
nemoclaw --helpThe exact version-check command was not confirmed in the current documentation excerpt, so do not rely on a command copied from an older post. Use the subcommands shown by nemoclaw --help and the version-specific CLI reference to inspect the installed release and status.
- Run
docker psif Docker is your runtime and confirm the NemoClaw containers remain up after onboarding. - Check
free -handdf -hafter images are pulled; a VPS that was barely above the minimum can run out of RAM or disk during normal operation. - Run a minimal agent interaction through the configured provider before adding web search, messaging or a public dashboard.
- Keep the API key out of command output and redact it from any logs you send to an administrator.
Troubleshoot the common failures
| Symptom | Likely cause | Fix |
|---|---|---|
| Installer rejects Node.js | Node is below 22.19, or the shell is using a different system Node. | Run node --version and which node. Install Node 22 through nvm, activate it with nvm use 22, verify npm 10 or later, then rerun the installer. |
nemoclaw: command not found |
Your current shell has not loaded the nvm/fnm path or the install directory. | Run source ~/.bashrc or source ~/.zshrc, or open a new terminal. Recheck the command in that same session. |
docker: permission denied |
The user is not yet in Docker’s group, or the group change has not reached the current shell. | Complete Docker’s user setup, run newgrp docker, then test docker ps. Do not run the NemoClaw installer as root just to hide the permission problem. |
| Docker daemon unavailable | The service is stopped, failed to start, or the VPS image lacks a working installation. | Start and inspect Docker using your distribution’s service tools, then rerun docker ps before retrying NemoClaw. |
| Out-of-memory kill during setup | Docker, k3s and the OpenShell gateway overlap; 8 GB can be marginal. | Move to a 16 GB VPS. Temporary swap may prevent an immediate kill but slows provisioning and is not equivalent to additional RAM. |
| Onboarding stopped halfway | SSH disconnect, invalid credential or a provider/model choice failed. | Correct the underlying issue and run nemoclaw onboard --resume. |
| Cloud inference authentication fails | The key is mistyped, expired, has whitespace, or belongs to a provider different from the selected one. | Re-enter the credential, confirm the nvapi- prefix and provider selection, and keep the key private. |
| Podman path fails immediately | NEMOCLAW_GATEWAY_RUNTIME=podman was set without a functioning rootless Podman setup. |
Test Podman as the NemoClaw user, or remove the variable and use the default Docker runtime. |
| Disk fills while pulling images | The VPS has only the 20 GB minimum and little headroom for layers and logs. | Free space or resize to the 40 GB recommended target before retrying. |
Security and operational choices after installation
Keep the management surface private
A successful install does not mean the gateway should be exposed directly to the public internet. Restrict SSH by source address where your VPS provider permits it, use key-based authentication, and open only the application ports you have deliberately configured. If you later add a dashboard or HTTPS proxy, place it behind the VPS firewall and authentication rather than exposing an administrative endpoint by default.
Separate first-run validation from integrations
The quickest way to diagnose a deployment is to validate one agent, one model and one provider first. Add web search and messaging only after that path works. Each integration introduces its own credentials, outbound network requirements and failure modes.
Plan for memory growth
RAM is the most important capacity decision for this stack. The 8 GB figure is a minimum, not a performance guarantee. Container restarts, model changes, logs and additional integrations can increase usage after installation. Monitor memory and disk before scheduling unattended jobs; resize the VPS rather than relying on swap when the system regularly approaches its limit.
Keeping the installation current
NemoClaw’s release line is changing quickly. The July 14, 2026 release notes are the latest dated source used for the Node minimum, and the 22.19 requirement applies from v0.0.82 onward. Check the current release notes and prerequisites before rebuilding a VPS or upgrading Node, Docker, the gateway or the agent. Do not treat an older third-party tutorial’s version numbers as authoritative.
Before an upgrade, record the selected runtime, provider, model and policy tier; back up any configuration or credentials according to your organization’s secret-management rules; and confirm that the VPS still has the recommended memory and disk headroom. If an upgrade leaves onboarding incomplete, use nemoclaw onboard --resume rather than creating a second installation state.
A repeatable deployment checklist
- VPS has 4 or more vCPUs, at least 8 GB RAM and 20 GB free disk; 16 GB RAM and 40 GB disk are preferred.
- Ubuntu version and NemoClaw support are confirmed against the current NVIDIA prerequisites.
node --versionreports 22.19 or newer.npm --versionreports 10 or newer.- Docker runs for the normal user, or rootless Podman is tested and selected intentionally.
- NVIDIA API key is available for cloud inference and stored securely.
- Official installer command is copied from the current documentation, not an old guide.
- For automation,
NEMOCLAW_NON_INTERACTIVE=1andNEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1are set before the installer command. - Onboarding selects OpenClaw, a provider, a model and the matching credential.
- Optional web search and messaging are deferred until the base deployment works.
- Suggested network-policy tier is reviewed and accepted.
nemoclaw --help, the runtime CLI and a minimal agent request work after onboarding.
With those checks complete, NemoClaw is running on a Linux VPS through its supported runtime, with enough capacity to avoid the most common setup failures and with the provider and policy choices recorded for future maintenance.
Free tools Windows power users keep installed
One-click scans. No signup required.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
