Installing and Running OpenClaw on gcube

This guide explains how to run the Ollama-based gemma4:e4b model and configure the OpenClaw Gateway and Control UI in the gcube workload environment.

Before You Begin

This guide proceeds in the following order:

Create and deploy workload
SSH access
Download Ollama gemma4:e4b model
Install Node.js and npm
Install OpenClaw and onboard
Install plugin dependencies
Run OpenClaw Gateway
Approve Telegram Pairing
Stop and verify OpenClaw process
Access web dashboard
Review and edit configuration file
Configure and use web dashboard
Troubleshooting
Security notes

Step 1 — Create and Deploy Workload

Register a workload with the following settings.

Item	Value
Registry Type	Docker Hub
Container Image	`ollama/ollama:latest`
Container Port	`18789`
GPU Selection	RTX Series
Other Options	Shared Memory 4GB

Note

Port 18789 is used for OpenClaw Gateway and Control UI access.

Workload creation screen

Step 2 — SSH Access

Once the workload is running, check the SSH connection information.

SSH connection info screen

Look up and register your public IP, then connect to the container using the SSH address and port.

ssh <SSH connection command provided by gcube>

Once PuTTY terminal opens successfully, log in with your username and password.

PuTTY connection screen

Tip

In PuTTY, you can paste using right-click.

Step 3 — Download Ollama Model

Enter the command below to download and run the gemma4:e4b model.

ollama run gemma4:e4b

gemma4 model download screen

Tip

In the Ollama chat screen, you can exit with the /bye command.

Step 4 — Install Node.js and npm

Install Node.js, npm, and git for OpenClaw installation.

apt-get update && apt-get install -y curl git
curl -fsSL https://deb.nodesource.com/setup_lts.x | bash -
apt-get install -y nodejs

Package update screen

After installation, verify everything is installed correctly.

node -v
npm -v
git --version

Step 5 — Install OpenClaw and Onboard

npm i -g openclaw
openclaw onboard

OpenClaw installation screen

OpenClaw onboard screen

Step 6 — Install Plugin Dependencies

If you see a plugin installation error during openclaw onboard, install the required dependencies with the commands below, then run onboard again.

npm i -g @larksuiteoapi/node-sdk nostr-tools
npm i -g @slack/web-api
npm i -g @whiskeysockets/baileys

openclaw onboard

OpenClaw onboard re-run screen

Warning

During onboarding, make sure to select the exact gemma4:e4b model under default model.

Step 7 — Run OpenClaw Gateway

Once onboarding is complete, run the Gateway.

openclaw gateway

Step 8 — Approve Telegram Pairing

When OpenClaw Gateway is running, a Telegram connection request will appear. Use the Pairing Code from Telegram to run the approval command.

openclaw pairing approve telegram <PAIRING_CODE>

Telegram Pairing screen

Warning

Keep the Pairing Code confidential and do not expose it externally.

Step 9 — Stop and Verify OpenClaw Process

If you need to restart the Gateway, stop the existing process first, then restart.

pkill -9 -f openclaw
ps -ef | grep openclaw
openclaw gateway

Step 10 — Access Web Dashboard

Access the OpenClaw Control UI via the gcube workload's service URL or port access URL.

http://<SERVICE_URL>:18789

Web dashboard access screen

Step 11 — Review and Edit Configuration File

Use the command below to write the /root/.openclaw/openclaw.json configuration file.

Warning

Make sure to replace <YOUR_GATEWAY_TOKEN>, <YOUR_TELEGRAM_BOT_TOKEN>, and <SERVICE_URL> with your actual values.

cat > /root/.openclaw/openclaw.json <<'EOF'
EOF

{
  "agents": {
    "defaults": {
      "workspace": "/root/.openclaw/workspace",
      "model": {
        "primary": "ollama/gemma4:e4b"
      },
      "models": {
        "ollama/gemma4:e4b": {}
      }
    }
  },
  "gateway": {
    "mode": "local",
    "auth": {
      "mode": "token",
      "token": "<YOUR_GATEWAY_TOKEN>"
    },
    "port": 18789,
    "bind": "lan",
    "controlUi": {
      "allowInsecureAuth": true,
      "allowedOrigins": [
        "http://<SERVICE_URL>:18789"
      ],
      "enabled": true,
      "dangerouslyDisableDeviceAuth": true
    },
    "tailscale": {
      "mode": "off",
      "resetOnExit": false
    },
    "nodes": {
      "denyCommands": [
        "camera.snap",
        "camera.clip",
        "screen.record",
        "contacts.add",
        "calendar.add",
        "reminders.add",
        "sms.send",
        "sms.search"
      ]
    },
    "tls": {
      "enabled": false,
      "autoGenerate": false
    }
  },
  "session": {
    "dmScope": "per-channel-peer"
  },
  "tools": {
    "profile": "coding",
    "web": {
      "search": {
        "provider": "ollama",
        "enabled": true
      }
    }
  },
  "plugins": {
    "entries": {
      "ollama": {
        "enabled": true
      },
      "telegram": {
        "enabled": true
      }
    }
  },
  "models": {
    "mode": "merge",
    "providers": {
      "ollama": {
        "baseUrl": "http://127.0.0.1:11434",
        "api": "ollama",
        "apiKey": "OLLAMA_API_KEY",
        "models": [
          {
            "id": "gemma4",
            "name": "gemma4",
            "reasoning": false,
            "input": ["text"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 128000,
            "maxTokens": 8192,
            "compat": { "supportsUsageInStreaming": true }
          },
          {
            "id": "gemma4:e4b",
            "name": "gemma4:e4b",
            "reasoning": true,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 131072,
            "maxTokens": 8192,
            "compat": {
              "supportsTools": true,
              "supportsUsageInStreaming": true
            }
          }
        ]
      }
    }
  },
  "auth": {
    "profiles": {
      "ollama:default": {
        "provider": "ollama",
        "mode": "api_key"
      }
    }
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "groups": {
        "*": { "requireMention": true }
      },
      "botToken": "<YOUR_TELEGRAM_BOT_TOKEN>"
    }
  },
  "wizard": {
    "lastRunAt": "2026-05-03T15:32:24.907Z",
    "lastRunVersion": "2026.5.2",
    "lastRunCommand": "onboard",
    "lastRunMode": "local"
  },
  "meta": {
    "lastTouchedVersion": "2026.5.2",
    "lastTouchedAt": "2026-05-03T15:32:25.021Z"
  }
}

Step 12 — Configure and Use Web Dashboard

Access the web dashboard via the gcube workload's service URL or port access URL.

http://<SERVICE_URL>:18789

From the OpenClaw web dashboard, you can review and update settings, and use OpenClaw directly through connected features.

OpenClaw web dashboard screen

Step 13 — Troubleshooting

Issue	Cause	Solution
Cannot access web dashboard	Gateway `bind` setting not configured for external access	Set `bind` to `lan` and `port` to `18789`
Control UI access blocked	`controlUi.allowedOrigins` does not match service URL	Add `http://<SERVICE_URL>:18789` to `allowedOrigins`
OpenClaw command not working	npm global install path issue	Re-run `npm i -g openclaw` and check PATH
Plugin error	Required npm package missing	Install the relevant package globally
Model fails to run	Ollama model not downloaded	Run `ollama run gemma4:e4b`
Pairing fails	Request ID or code mismatch	Re-check the latest pairing request

Note

OpenClaw is frequently updated, so commands, configuration fields, and UI may change. If something differs from this guide, check the official OpenClaw documentation first.

Step 14 — Security Notes

Warning

Do not expose the Gateway Token publicly.
Do not enter the Telegram Bot Token directly in documents.
Remove the Pairing Code and Request ID from documents after use.
Only enter http://<SERVICE_URL>:18789 in allowedOrigins.
allowInsecureAuth: true, dangerouslyDisableDeviceAuth: true, and tls.enabled: false settings are recommended for test environments only.

gcube CLI → Use Workloads →