Security Hardening for OpenClaw on Azure

Why Security Matters for AI Gateways

OpenClaw sits between your AI models and your messaging platforms. A compromised gateway means:

• Unauthorized access to your AI model API keys
• Ability to impersonate your bot on connected channels
• Access to conversation history and user data
• Potential API cost exposure

OpenClaw takes security seriously with multiple layers of protection. Let’s configure them properly for Azure.

Layer 1: Gateway Authentication Token

The gateway token is the first line of defense. It’s auto-generated during docker-setup.sh and stored in .env:

cd ~/openclaw
grep OPENCLAW_GATEWAY_TOKEN .env

Best practices

• Never share the token in plain text
• Rotate it periodically:

  # Generate a new token
  NEW_TOKEN=$(openssl rand -hex 32)
  echo "New token: $NEW_TOKEN"
  
  # Update .env
  sed -i "s/OPENCLAW_GATEWAY_TOKEN=.*/OPENCLAW_GATEWAY_TOKEN=$NEW_TOKEN/" .env
  
  # Restart
  docker compose down
  docker compose up -d

• Use a strong token — at least 32 hex characters

Layer 2: Device Pairing

OpenClaw requires browsers to pair with the gateway before accessing the Control UI. This creates a device identity tied to the browser’s secure context.

How device pairing works

1. Browser requests access → generates a device identity (requires HTTPS or localhost)
2. Gateway stores the request as “pending”
3. Operator approves via CLI
4. Browser is paired and can authenticate

Managing devices

# List all paired and pending devices
docker compose run --rm openclaw-cli devices list

# Approve a pending device
docker compose run --rm openclaw-cli devices approve <requestId>

# Revoke a device (if supported)
docker compose run --rm openclaw-cli devices list

Break-glass access

If you’re locked out and can’t pair (e.g., no HTTPS, no localhost):

# Generate a direct dashboard URL with token
docker compose run --rm openclaw-cli dashboard --no-open

Warning: The following flags should only be used temporarily for debugging:

• gateway.controlUi.allowInsecureAuth — allows token-only auth over HTTP
• gateway.controlUi.dangerouslyDisableDeviceAuth — disables device checks entirely

Layer 3: Network Security (Azure NSG)

Azure Network Security Groups (NSGs) are your perimeter defense.

Recommended NSG rules

Setting up via Azure CLI

# Allow SSH from your IP only
az network nsg rule create \
  --resource-group rg-openclaw \
  --nsg-name vm-openclaw-01NSG \
  --name AllowSSH \
  --priority 1000 \
  --destination-port-ranges 22 \
  --source-address-prefixes "<YOUR_IP>" \
  --protocol Tcp --access Allow

# Allow OpenClaw from your IP only (skip if using SSH tunnel)
az network nsg rule create \
  --resource-group rg-openclaw \
  --nsg-name vm-openclaw-01NSG \
  --name AllowOpenClaw \
  --priority 1001 \
  --destination-port-ranges 18789 18790 \
  --source-address-prefixes "<YOUR_IP>" \
  --protocol Tcp --access Allow

If using SSH tunnel only

You don’t need the OpenClaw NSG rule at all — just allow SSH:

OPENCLAW_GATEWAY_BIND=loopback  # in .env

Layer 4: CORS Origin Enforcement

When the gateway binds to non-loopback, OpenClaw enforces browser origin checks.

Explicit origins (recommended)

docker compose run --rm openclaw-cli config set \
  gateway.controlUi.allowedOrigins \
  '["https://openclaw.yourdomain.com"]'

Dangerous flags to avoid in production

Layer 5: HTTPS with Tailscale (Optional but Recommended)

For production deployments with remote access, HTTPS is the gold standard. Tailscale Serve is the easiest way to add HTTPS to OpenClaw:

Install Tailscale on the VM

curl -fsSL https://tailscale.com/install.sh | sh
sudo tailscale up

Expose OpenClaw via Tailscale Serve

sudo tailscale serve https / http://127.0.0.1:18789

Now access the Control UI at:

https://<your-machine-name>.<your-tailnet>.ts.net/

Benefits:

• Automatic HTTPS certificates
• No public ports needed
• Device authentication works properly (secure context)
• Access from any Tailscale-connected device

Layer 6: Built-in Security Audit

OpenClaw includes a security audit command that checks your configuration:

docker compose run --rm openclaw-cli security audit

This command checks for:

Example audit output

The audit will report findings with severity levels and remediation suggestions. Address all critical findings before going to production.

Security Checklist

Before running OpenClaw in production on Azure:

• Gateway token is strong (32+ hex chars) and stored securely
• .env file has restricted permissions (chmod 600 .env)
• NSG rules restrict access to your IP(s) only
• Bind mode is loopback (SSH tunnel) or lan with explicit allowedOrigins
• No dangerous* flags enabled in production
• Device pairing is active (not disabled)
• security audit reports no critical findings
• Bot tokens (Discord, etc.) are valid and not exposed
• HTTPS via Tailscale Serve or reverse proxy (for remote access)
• Regular token rotation schedule established

Next Steps

Now let’s look at common deployment issues: Troubleshooting OpenClaw Docker Deployments.