### Example API Key Export for Non-Interactive Setup

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart.mdx

Export the API key environment variable before launching the installer for a non-interactive setup. This avoids the wizard prompting for the key.

```bash
export NVIDIA_API_KEY=<your-key>
```

--------------------------------

### Run Onboarding Wizard

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands-nemohermes.mdx

Initiate the interactive setup wizard for new installations. This command creates an OpenShell gateway, registers inference providers, builds the sandbox image, and creates the sandbox. It is recommended for new installs and for recreating a sandbox after configuration changes.

```bash
nemohermes onboard [--non-interactive] [--resume | --fresh] [--recreate-sandbox] [--gpu | --no-gpu] [--from <Dockerfile>] [--name <sandbox>] [--sandbox-gpu | --no-sandbox-gpu] [--sandbox-gpu-device <device>] [--agent <name>] [--control-ui-port <N>] [--yes | -y] [--no-ollama-autostart] [--yes-i-accept-third-party-software]
```

--------------------------------

### Run Interactive Onboarding Wizard

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

Use the `onboard` command to run the interactive setup wizard for new installations. It configures the OpenShell gateway, inference providers, and sandbox environment. This command is recommended for initial setup and after significant configuration changes.

```bash
$$nemoclaw onboard [--non-interactive] [--resume | --fresh] [--recreate-sandbox] [--gpu | --no-gpu] [--from <Dockerfile>] [--name <sandbox>] [--sandbox-gpu | --no-sandbox-gpu] [--sandbox-gpu-device <device>] [--agent <name>] [--control-ui-port <N>] [--yes | -y] [--no-ollama-autostart] [--yes-i-accept-third-party-software]
```

--------------------------------

### Install NemoClaw and Run Onboard Wizard

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/index.mdx

This command installs NemoClaw and initiates the onboard wizard. It's the primary method to get started with the NemoClaw stack.

```shell
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
```

--------------------------------

### Install or Start Managed vLLM

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/inference/inference-options.mdx

Use NEMOCLAW_PROVIDER=install-vllm for non-interactive runs to have NemoClaw install or start a managed vLLM container. Add NEMOCLAW_EXPERIMENTAL=1 on generic Linux NVIDIA GPU hosts.

```bash
NEMOCLAW_PROVIDER=install-vllm \
  $$nemoclaw onboard --non-interactive
```

--------------------------------

### Install All Dependencies

Source: https://github.com/nvidia/nemoclaw/blob/main/AGENTS.md

Installs all project dependencies, links the main package, builds the plugin, and synchronizes blueprint dependencies.

```bash
npm install && npm link && cd nemoclaw && npm install && npm run build && cd .. && cd nemoclaw-blueprint && uv sync && cd ..
```

--------------------------------

### Install binutils on Debian/Ubuntu

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/prerequisites.mdx

If the installer reports that 'strings' is missing, install 'binutils' using apt-get and rerun the installer. This ensures the OpenShell binary can be verified.

```bash
sudo apt-get install -y binutils
```

--------------------------------

### Non-Interactive Install and Onboarding

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart.mdx

Run installation and onboarding without prompts by setting non-interactive mode and provider variables.

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_NON_INTERACTIVE=1 NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1 bash
```

--------------------------------

### Non-Interactive Hermes Setup

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-get-started/references/quickstart-hermes.md

For CI or scripted installations, set the required environment variables before running the installer. This example configures NVIDIA Endpoints and names the sandbox 'my-hermes'.

```bash
$ export NEMOCLAW_AGENT=hermes
$ export NEMOCLAW_NON_INTERACTIVE=1
$ export NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1
$ export NEMOCLAW_SANDBOX_NAME=my-hermes
$ export NVIDIA_API_KEY=<your-key>
$ curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
```

--------------------------------

### Action Examples Directory

Source: https://github.com/nvidia/nemoclaw/blob/main/src/lib/actions/README.md

Provides examples of action file names and their corresponding workflow orchestrations.

```text
dns/index.ts                    # internal dns fix-coredns/setup-proxy orchestration
dev/npm-link-or-shim.ts          # prepare-time dev shim orchestration
installer/plan.ts                # deterministic installer planning
uninstall/plan.ts                # host uninstall planning
uninstall/run-plan.ts            # uninstall plan application
sandbox/*.ts                     # public sandbox workflows
```

--------------------------------

### Start vLLM Server

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/inference/inference-options.mdx

Starts a vLLM model server. This is an example for an OpenAI-compatible server that NemoClaw can connect to.

```bash
vllm serve meta-llama/Llama-3.1-8B-Instruct --port 8000
```

--------------------------------

### Run Windows Bootstrap Script for NemoClaw Setup

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/windows-preparation.mdx

This PowerShell script automates the setup of WSL 2, Ubuntu, and Docker Desktop on Windows. It handles feature enablement, installation, and configuration, including Docker Desktop's WSL integration. Run this from Windows PowerShell.

```powershell
Invoke-WebRequest -Uri 'https://raw.githubusercontent.com/NVIDIA/NemoClaw/main/scripts/bootstrap-windows.ps1' -OutFile "$env:TEMP\bootstrap-windows.ps1"; powershell.exe -ExecutionPolicy Bypass -File "$env:TEMP\bootstrap-windows.ps1"
```

--------------------------------

### Run Interactive Sandbox Setup Wizard

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-reference/references/commands.md

Initiates the interactive setup wizard for creating and configuring a NemoClaw sandbox. Supports options for non-interactive setup, resuming, recreating the sandbox, GPU configuration, specifying a Dockerfile, naming the sandbox, and agent selection.

```bash
nemoclaw onboard [--non-interactive] [--resume | --fresh] [--recreate-sandbox] [--gpu | --no-gpu] [--from <Dockerfile>] [--name <sandbox>] [--sandbox-gpu | --no-sandbox-gpu] [--sandbox-gpu-device <device>] [--agent <name>] [--control-ui-port <N>] [--yes | -y] [--no-ollama-autostart] [--yes-i-accept-third-party-software]
```

--------------------------------

### Example Dockerfile for Baking OpenClaw Plugins

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/deployment/install-openclaw-plugins.mdx

This Dockerfile copies a plugin into the sandbox, installs its dependencies, builds it, and registers it with OpenClaw. Ensure the base image is compatible with your plugin.

```dockerfile
ARG SANDBOX_BASE=ghcr.io/nvidia/nemoclaw/sandbox-base:latest
FROM ${SANDBOX_BASE}

COPY my-plugin/ /opt/my-plugin/
WORKDIR /opt/my-plugin
RUN npm ci --no-audit --no-fund && npm run build

RUN mkdir -p /sandbox/.openclaw/extensions \
 && cp -a /opt/my-plugin /sandbox/.openclaw/extensions/my-plugin \
 && openclaw doctor --fix

WORKDIR /opt/nemoclaw
```

--------------------------------

### Prepare for OpenShell Upgrade during Installation

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

Set this environment variable to allow the installer to automatically handle OpenShell upgrades, including backing up sandbox state, retiring the old gateway, and restoring state. This is for scripted installs.

```bash
NEMOCLAW_ACCEPT_EXPERIMENTAL_OPENSHELL_UPGRADE=1
```

--------------------------------

### Accept Third-Party Software with Installer Flag

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart.mdx

Pass the installer flag through bash -s for accepting third-party software.

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash -s -- --yes-i-accept-third-party-software
```

--------------------------------

### Start Colima on macOS

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/troubleshooting.mdx

Start the Colima container runtime on a macOS host.

```bash
colima start
```

--------------------------------

### Non-interactive vLLM Server Setup

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/inference/inference-options.mdx

Onboard with an already-running vLLM server using the NEMOCLAW_PROVIDER=vllm environment variable for non-interactive setup.

```bash
NEMOCLAW_PROVIDER=vllm \
  $$nemoclaw onboard --non-interactive
```

--------------------------------

### Install Homebrew Formula

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-manage-policy/references/integration-policy-examples.md

Use this command to install a Homebrew formula within the sandbox environment. The `brew` preset must be applied beforehand.

```console
$ nemoclaw my-assistant policy-add brew --yes
$ nemoclaw my-assistant exec -- brew install <formula>
```

--------------------------------

### Internal Installer Plan Command

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

Builds a deterministic installer plan from environment and probe inputs without applying it. Used by install.sh.

```bash
$$nemoclaw internal installer plan
```

--------------------------------

### Install NemoClaw with Hermes Agent

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart-hermes.mdx

Set the NEMOCLAW_AGENT environment variable to 'hermes' and then run the installer script. This installs the CLI, selects the 'nemohermes' alias, and starts the guided onboarding flow.

```bash
export NEMOCLAW_AGENT=hermes
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
```

--------------------------------

### Serve Documentation Locally

Source: https://github.com/nvidia/nemoclaw/blob/main/AGENTS.md

Starts a local server to preview the project documentation.

```bash
npm run docs:live
```

--------------------------------

### Onboard Multiple Sandboxes

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/troubleshooting.mdx

Demonstrates onboarding multiple sandboxes, showing the default port assignment and explicit port override for distinct dashboards.

```bash
$$nemoclaw onboard
$$nemoclaw onboard
$$nemoclaw onboard --control-ui-port 19000
```

--------------------------------

### Non-Interactive Hermes Setup with NVIDIA Endpoints

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart-hermes.mdx

For automated installations, set environment variables before running the NemoClaw installer script. This example configures Hermes for NVIDIA Endpoints and names the sandbox 'my-hermes'.

```bash
export NEMOCLAW_AGENT=hermes
export NEMOCLAW_NON_INTERACTIVE=1
export NEMOCLAW_ACCEPT_THIRD_PARTY_SOFTWARE=1
export NEMOCLAW_SANDBOX_NAME=my-hermes
export NVIDIA_API_KEY=<your-key>
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
```

--------------------------------

### Start Fresh Onboarding with New Choices

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/troubleshooting.mdx

Discard a previous failed onboarding session and start over with new provider and model selections. Use this when the original failure was due to incorrect choices.

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash -s -- --fresh
```

--------------------------------

### Homebrew Installation and Execution

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/network-policy/integration-policy-examples.mdx

Adds the 'brew' preset, installs a Homebrew formula, and executes a formula command within the sandbox. Assumes Homebrew is already set up.

```bash
$$nemoclaw my-assistant policy-add brew --yes
$$nemoclaw my-assistant exec -- brew install <formula>
$$nemoclaw my-assistant exec -- bash -lc '<formula-command>'
```

--------------------------------

### Install NemoClaw with Hermes Agent

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-get-started/references/quickstart-hermes.md

Install the NemoClaw CLI and start the Hermes onboarding flow. Ensure the NEMOCLAW_AGENT environment variable is set to 'hermes'.

```bash
$ export NEMOCLAW_AGENT=hermes
$ curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
```

--------------------------------

### Start Host Auxiliary Services (Cloudflare Tunnel)

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands-nemohermes.mdx

Starts optional host auxiliary services, primarily the Cloudflare tunnel if cloudflared is installed. This exposes the dashboard with a public URL. For named tunnels, set the CLOUDFLARE_TUNNEL_TOKEN environment variable.

```bash
nemohermes tunnel start
```

```bash
export CLOUDFLARE_TUNNEL_TOKEN=<cloudflare-tunnel-token>
nemohermes tunnel start
```

--------------------------------

### Onboard with Model Router

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-configure-inference/references/inference-options.md

Use this command to onboard with the Model Router when using a scripted setup. Ensure you have your NVIDIA API key.

```bash
$ NEMOCLAW_PROVIDER=routed NVIDIA_API_KEY=<your-key> nemoclaw onboard --non-interactive
```

--------------------------------

### Manual OpenShell Gateway and Sandbox Preparation

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

These commands are used for manual preparation before running the installer when automatic OpenShell upgrades are not desired or supported. They involve backing up, removing, and destroying the old gateway, and stopping any running host gateway.

```bash
$$nemoclaw backup-all
```

```bash
openshell gateway remove nemoclaw || openshell gateway destroy -g nemoclaw || openshell gateway destroy
```

```bash
sudo pkill -f openshell-gateway
```

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_OPENSHELL_UPGRADE_PREPARED=1 bash
```

--------------------------------

### Onboard Hermes Agent with NemoClaw

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart-hermes.mdx

If NemoClaw is already installed, you can start the Hermes onboarding process directly using the 'nemohermes onboard' command.

```bash
nemohermes onboard
```

--------------------------------

### Build Documentation

Source: https://github.com/nvidia/nemoclaw/blob/main/AGENTS.md

Builds the project's documentation.

```bash
npm run docs
```

--------------------------------

### Build Sandbox from Custom Dockerfile (Alternative Path)

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands-nemohermes.mdx

An alternative example demonstrating the use of `--from` with a custom Dockerfile named differently, such as `Dockerfile.custom`.

```bash
nemohermes onboard --from ./Dockerfile.custom
```

--------------------------------

### Manage OpenShell Dashboard Port Forward

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-manage-sandboxes/SKILL.md

Starts or lists active port forwards for the OpenShell dashboard. Requires the dashboard port from the install summary.

```console
$ openshell forward start --background <dashboard-port> my-gpt-claw
```

```console
$ openshell forward list
```

--------------------------------

### Deprecated Setup Spark Command

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

This command is deprecated and serves as a compatibility alias for `$$nemoclaw onboard`. Use the standard installer and `$$nemoclaw onboard` instead.

```bash
$$nemoclaw setup-spark
```

--------------------------------

### Example Build Directory Structure

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/deployment/install-openclaw-plugins.mdx

Organize your Dockerfile and plugin source code in a dedicated directory. The parent directory of the Dockerfile serves as the Docker build context.

```text
my-plugin-sandbox/
├── Dockerfile
└── my-plugin/
    ├── package.json
    └── src/
```

--------------------------------

### Live Scenario Test with NemoClaw Fixtures

Source: https://github.com/nvidia/nemoclaw/blob/main/test/e2e-scenario/docs/README.md

Example of a Vitest test using NemoClaw fixtures to define and test a scenario. Fixtures abstract away setup and teardown complexities.

```typescript
import { test } from "../framework/e2e-test.ts";

test("ubuntu repo cloud OpenClaw", async ({ 
  repo,
  openclaw,
  gateway,
  sandbox,
  inference,
}) => {
  await repo.installCurrent();

  const instance = await openclaw.onboard({
    agent: "openclaw",
    provider: "nvidia",
  });

  await gateway.expectHealthy(instance);
  await sandbox.expectRunning(instance);
  await inference.expectLocalChat(instance, { prompt: "Say ok.", expect: /ok/i });
});
```

--------------------------------

### Set Gateway Bind Address and Port for Onboarding

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands-nemohermes.mdx

Example of setting both the gateway bind address and port environment variables before onboarding a service.

```bash
NEMOCLAW_GATEWAY_BIND_ADDRESS=0.0.0.0 NEMOCLAW_GATEWAY_PORT=8990 nemohermes onboard
```

--------------------------------

### Install a NemoClaw Skill

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-reference/references/commands.md

Deploys a skill directory to a running sandbox. The command validates the SKILL.md frontmatter, uploads non-dot files, and performs agent-specific post-install steps. Files starting with '.' are skipped.

```bash
nemoclaw my-assistant skill install ./my-skill/
```

--------------------------------

### Quick Start Chat UI

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-deploy-remote/SKILL.md

Connect to an existing NemoClaw assistant and launch the OpenClaw TUI. Assumes the Brev instance is already provisioned and onboarded.

```console
$ nemoclaw my-assistant connect
$ openclaw tui
```

--------------------------------

### Get Current Inference Provider and Model

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/inference/switch-inference-providers.mdx

Use this command to display the provider and model the gateway is currently routing inference requests to. Run it before setting a new route to confirm the starting state, or after a switch to verify the change.

```bash
$$nemoclaw inference get
```

--------------------------------

### Install Xcode Command Line Tools on macOS

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/troubleshooting.mdx

On macOS, install Xcode Command Line Tools before running the NemoClaw installer to ensure compatibility with the installer and Node.js toolchain.

```bash
xcode-select --install
```

--------------------------------

### Action File Structure Example

Source: https://github.com/nvidia/nemoclaw/blob/main/src/lib/actions/README.md

Illustrates the preferred file layout for actions within the project, separating workflow logic into distinct files.

```text
src/lib/actions/<area>/<verb>.ts
src/lib/actions/<area>/<verb>.test.ts
```

--------------------------------

### Set Ollama Install Mode to System

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands-nemohermes.mdx

Configure NEMOCLAW_OLLAMA_INSTALL_MODE to 'system' to use the official installer script, which installs to /usr/local and sets up systemd.

```bash
export NEMOCLAW_OLLAMA_INSTALL_MODE=system
```

--------------------------------

### Create Snapshot and Update Sandboxes

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/manage-sandboxes/lifecycle.mdx

Perform a manual upgrade workflow by creating a snapshot before updating. This includes creating a pre-upgrade snapshot, running the general update command, and then checking for sandbox upgrades.

```bash
$$nemoclaw <sandbox-name> snapshot create --name pre-upgrade
$$nemoclaw update --yes
$$nemoclaw upgrade-sandboxes --check
```

--------------------------------

### Install Dependencies and Build TypeScript Plugin

Source: https://github.com/nvidia/nemoclaw/blob/main/CONTRIBUTING.md

Installs root dependencies and builds the TypeScript plugin for NemoClaw. Ensure Node.js and npm are installed.

```bash
# Install root dependencies (OpenClaw + CLI entry point)
npm install

# Install and build the TypeScript plugin
cd nemoclaw && npm install && npm run build && cd ..

# Install Python deps for the blueprint
cd nemoclaw-blueprint && uv sync && cd ..

```

--------------------------------

### Review Configuration Summary

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-get-started/SKILL.md

This text output shows a summary of the configuration before the sandbox build. It includes provider, model, API key status, and sandbox name. Press 'Y' to confirm or 'n' to abort.

```text
  ──────────────────────────────────────────────────
  Review configuration
  ──────────────────────────────────────────────────
  Provider:      compatible-endpoint
  Model:         openai/openai/gpt-5.5
  API key:       COMPATIBLE_API_KEY (staged for OpenShell gateway registration)
  Web search:    disabled
  Messaging:     none
  Sandbox name:  my-gpt-claw
  Note:          Sandbox build typically takes 5–15 minutes on this host.
  ──────────────────────────────────────────────────
  Web search and messaging channels will be prompted next.
  Apply this configuration? [Y/n]:
```

--------------------------------

### Internal Installer Normalize Environment Command

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

Normalizes installer ref and provider environment values without applying installation changes. Used by install.sh.

```bash
$$nemoclaw internal installer normalize-env
```

--------------------------------

### Sub-Agent Delegation Instructions Example

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-configure-inference/references/set-up-sub-agent.md

Place task delegation instructions in the TOOLS.md file within the writable workspace. This example shows a typical path for such a file.

```text
/sandbox/.openclaw/workspace/TOOLS.md
```

--------------------------------

### Onboarding with Custom Name Suggestion

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

The CLI suggests a valid slug format for sandbox names that do not conform to the rules. Use --name to provide a sandbox name.

```bash
$$nemoclaw --name MyAssistant
```

--------------------------------

### Install Ollama inside WSL

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/windows-preparation.mdx

Installs Ollama within the WSL environment using the official installation script. This allows NemoClaw to use Ollama for local inference.

```bash
curl -fsSL https://ollama.com/install.sh | sh
```

--------------------------------

### Prepare for OpenShell Upgrade during Installation

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands-nemohermes.mdx

This command prepares for an OpenShell upgrade during NemoClaw installation by allowing the installer to back up sandbox state, retire the old gateway, install the new OpenShell release, and restore state. Set this environment variable to enable the automatic upgrade path.

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_ACCEPT_EXPERIMENTAL_OPENSHELL_UPGRADE=1 bash
```

--------------------------------

### Create a snapshot of the workspace

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-reference/references/troubleshooting.md

Before running onboard, back up your workspace by creating a snapshot. This ensures your workspace files are protected.

```bash
nemoclaw <name> snapshot create
```

--------------------------------

### Onboard Custom Sandbox with Dockerfile

Source: https://github.com/nvidia/nemoclaw/blob/main/skills/nemoclaw-user-deploy-remote/references/install-openclaw-plugins.md

Command to onboard a custom sandbox by pointing to the Dockerfile in the build directory.

```console
$ nemoclaw onboard --from ./my-plugin-sandbox/Dockerfile
```

--------------------------------

### Install NemoClaw CLI

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/commands.mdx

Installs the NemoClaw CLI. Use the appropriate variant for OpenClaw or Hermes.

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash
```

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | NEMOCLAW_AGENT=hermes bash
```

--------------------------------

### Opt-in Live Vitest Scenarios

Source: https://github.com/nvidia/nemoclaw/blob/main/test/e2e-scenario/docs/MIGRATION.md

Build the CLI and then run live E2E scenarios using Vitest by setting the appropriate environment variable.

```bash
npm run build:cli
```

```bash
NEMOCLAW_RUN_E2E_SCENARIOS=1 npx vitest run --project e2e-scenarios-live --silent=false --reporter=default
```

--------------------------------

### Review Configuration Summary for OpenAI-Compatible Endpoint

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/get-started/quickstart.mdx

This is an example of the configuration review summary printed by the NemoClaw wizard before building a sandbox. It displays the selected provider, model, and other settings. Press 'Y' to confirm or 'n' to abort.

```text
  ──────────────────────────────────────────────────
  Review configuration
  ──────────────────────────────────────────────────
  Provider:      compatible-endpoint
  Model:         openai/openai/gpt-5.5
  API key:       configured for OpenShell gateway registration
  Web search:    disabled
  Managed tools: none
  Messaging:     none
  Sandbox name:  my-gpt-claw
  Note:          Sandbox build typically takes 5–15 minutes on this host.
  ──────────────────────────────────────────────────
  Web search and messaging channels will be prompted next.
  Apply this configuration? [Y/n]:
```

--------------------------------

### Start OpenShell gateway and resume onboarding

Source: https://github.com/nvidia/nemoclaw/blob/main/docs/reference/troubleshooting.mdx

If the 'connect' command fails because the gateway is down, you can start the gateway and resume the onboarding process. This ensures the gateway is running before attempting to connect or forward dashboards.

```bash
openshell gateway start --name nemoclaw
$$nemoclaw onboard --resume
$$nemoclaw <name> connect
```