How to Deploy an OpenClaw Shopping Assistant

An OpenClaw shopping assistant handles your customer interactions automatically.

It answers product questions at any hour, tracks orders instantly, and manages routine support tasks so your team doesn’t have to.

For your business, this means 24/7 coverage for product inquiries and order handling.

Your support team spends less time answering the same questions over and over, which cuts down your support costs and your customers get immediate answers, so they’re happier with the experience.

By the end of this guide, you’ll have a fully working shopping assistant deployed on your own server.

It will connect to your messaging channels, pull from your product catalog, and talk to customers in your brand’s voice.

Prerequisites

Before you start, get these:

• Cloud Server (VPS): Minimum 2 CPU cores and 4 GB RAM. Anything less will struggle with real-time conversations and headless browser tasks.
• Static or Elastic IP: Required for production. A consistent IP keeps your channel connections (WhatsApp, Telegram, etc.) stable. A changing IP causes authentication failures.
• LLM API Key: Get one from a provider like DeepSeek-V3 (budget-friendly for most e-commerce), GPT-5, or Claude. Start with a cheaper option and upgrade only if needed.
• Messaging Platform Accounts: Bot tokens or account access for your chosen platforms. For Telegram, create a bot via @BotFather. For WhatsApp, you’ll scan a QR code.

Step 1: Provision Server

Start by selecting a VPS provider and region. Truehost offers plans with dedicated resources and one-click OpenClaw deployment, making this step straightforward.

If you’re serving customers in Nigeria, choose a local region. For instance, Truehost gives you better latency and faster response times for your shoppers because we are local.

Once you’ve chosen your plan, attach a static IP immediately. Don’t delay this. Some providers allow you to reserve an IP before provisioning. Do that.

It will save you headaches down the line when you’re setting up channel connections.

Step 2: Install Prerequisites

SSH into your server and install the necessary tools:

bash
# Update package lists
sudo apt update && sudo apt upgrade -y

# Install Node.js (Node 24 is the recommended target)
curl -fsSL https://deb.nodesource.com/setup_24.x | sudo -E bash -
sudo apt install -y nodejs

# Install npm, git, and lsof
sudo apt install -y npm git lsof

# Verify installations
node --version
git --version

The OpenClaw installer script handles Node installation automatically if needed, but having it in place upfront prevents issues.

Step 3: Install OpenClaw

Run the official installer. This command detects your operating system, installs Node if needed, installs OpenClaw, and launches onboarding:

bash
curl -fsSL https://openclaw.ai/install.sh | bash

Note: Depending on your version, the command may be either openclaw or clawdbot. If one doesn’t work, try the other.

If you want to skip the onboarding step for now, use:

bash
curl -fsSL https://openclaw.ai/install.sh | bash -s -- --no-onboard

After installation, verify it worked by checking the version:

bash
openclaw --version

If the command isn’t found in a new terminal, you may need to source your profile or check the Node.js installation path.

Step 4: Run Onboarding Wizard

The onboarding wizard is your shortcut to setup. Run it:

bash
openclaw onboard

The wizard will guide you through several key decisions:

Model Provider + API Key
You’ll select your LLM provider and enter your API key. The wizard stores this securely. Don’t hardcode keys in scripts or configuration files. Always use the console panel or environment variables.

Gateway Configuration
The gateway is the central control plane for channels, sessions, and agents. The wizard will configure it to run as a background service. The default port is 18789.

If you have an existing OpenClaw installation that needs repair or migration, run:

bash
openclaw onboard --install-daemon
openclaw gateway install
openclaw doctor

The doctor command checks your configuration and health.

Step 5: Install Shopping Skills

Skills extend OpenClaw with specific capabilities. For a shopping assistant, you’ll install commerce-related skills.

The primary shopping assistant skill is called “Mai”. It handles product discovery, price comparison, stock management, and order requests.

It can run local-only for a single agent’s private catalog or registry-backed when multiple agents need to share data.

Install the Mai skill:

bash
clawhub --workdir ~/.openclaw/workspace --dir skills install mai
openclaw plugins install clawhub:mai-plugin

For platform-specific shopping skills, the installation depends on the platforms you support:

bash
# For a generic Shopify-style setup
openclaw plugins install shopify-skill

# For specific regional platforms (Zepto, etc.)
# Check the skill repository for exact names

After installation, configure skill-specific settings like API keys and store paths. For Mai, you can create a merchant catalog using the CLI helper:

bash
python3 scripts/mai.py merchant create --id your-store --name "Your Store Name" --city YourCity --contact "wechat:yourhandle" --tags "products,tags"
python3 scripts/mai.py product add --merchant your-store --sku product-001 --title "Product Name" --price 99 --stock 10 --category "category"

The default data path for Mai is ~/.local/share/mai/mai.json. You can specify a different path with --data.

Step 6: Configure Shopping Persona

This step is where you give your assistant personality. The workspace directory ~/.openclaw/workspace/ contains files that define who your assistant is, who it serves, and how it works.

Edit SOUL.md – This is the assistant’s personality and boundaries. It defines tone, style, and behavior principles .

Open the file:

bash
nano ~/.openclaw/workspace/SOUL.md

A shopping assistant example:

text
# SOUL.md - Who You Are

You are a shopping assistant. You help customers find products, compare prices, and make orders. You're helpful but not pushy.

## Behavior Principles
- Be neutral between buyer and merchant. Explain options, prices, and inventory without fabricating availability [citation:1].
- Confirm before irreversible steps. Ask for customer confirmation before creating an order.
- Surface risk plainly: low stock, missing merchant contact, unclear payment terms.
- Prefer short, actionable answers: recommendation, reason, risk, next action.
- Skip the "Great question!" fluff. Just help.

## Boundaries
- Never claim payment success without evidence.
- Don't hold funds directly.
- Always disclose when you're unsure.

Edit USER.md – This tells the assistant about you and your customers.

bash
nano ~/.openclaw/workspace/USER.md

Example:

text
# USER.md - Who We Serve

## Store Profile
- Store Name: Your Store
- Products: [Your product categories]
- Shipping: [X-Y] business days, free over $[amount]
- Returns: [X] days, unused items

## Customer Profile
- Primary audience: [Your customer description]
- Preferred tone: Friendly, concise, solution-oriented
- Timezone: [Your primary timezone]

Edit AGENTS.md -This defines the workflow for shopping tasks.

bash
nano ~/.openclaw/workspace/AGENTS.md

Example:

text
# AGENTS.md - Workflow Rules

## Shopping Task Flow
1. Identify customer intent (question, order, complaint)
2. For product questions: Search catalog, return top matches
3. For order status: Ask for order number, query system
4. For complaints: Empathize, offer solution, escalate if unresolved in 2 messages

## Escalation Rule
If you cannot resolve an issue in 2 messages:
"Let me connect you with our specialist. Please share your order number."

## Safety Rules
- Don't create orders without confirmation
- Don't share other customers' data
- Don't modify inventory without explicit confirmation

These three files work together: SOUL.md sets the personality, USER.md describes the context, and AGENTS.md defines the process.

Step 7: Connect Messaging Channels

Your shopping assistant needs to reach customers where they are. Here’s how to set up the major platforms.

Telegram:

• Open Telegram and search for @BotFather
• Send /newbot and follow the prompts
• Save the bot token you receive
• In OpenClaw, configure the channel:

bash
openclaw configure
# Select Telegram and paste your bot token

WhatsApp:

WhatsApp requires QR code scanning and pairing approval. The setup varies by version:

• Run the gateway to generate a QR code
• Scan it with WhatsApp on your phone
• Approve the pairing
• The assistant will then be available on WhatsApp Web

WeChat / QQ (Domestic Platforms):
For Chinese platforms, setup typically involves registering an official account and obtaining API credentials from Tencent. The process is platform-specific and may require additional approvals.

Important: Don’t set up five channels at once. Pick one, your highest-traffic platform, and get it working perfectly first. Then expand to secondary platforms.

Step 8: Enable Daemon & Test

Now you need to keep your assistant running. OpenClaw installs as a systemd user service on Linux.

Enable and start the gateway service:

bash
systemctl --user enable openclaw-gateway
systemctl --user start openclaw-gateway

Check that it’s running:

bash
systemctl --user status openclaw-gateway

To ensure the service persists across reboots:

bash
loginctl enable-linger $(whoami)
export XDG_RUNTIME_DIR=/run/user/$(id -u)

For manual control:

bash
# Start
systemctl --user start openclaw-gateway

# Stop
systemctl --user stop openclaw-gateway

# Restart
systemctl --user restart openclaw-gateway

Send a test message through your configured channel. Ask a product question. Check if you get a response to verify the whole pipeline is working.

If the gateway refuses to bind, you may need to set the gateway mode to bind to all interfaces instead of just localhost.

Step 9: Backup & Maintain

Your configuration and customer data are valuable. Back them up.

Backup the workspace directory:

bash
cp -r ~/.openclaw/workspace ~/backups/openclaw-workspace-$(date +%Y%m%d)

Use OpenClaw’s backup command:

bash
openclaw backup create

This creates a backup of your configuration, skills, and session data. You can restore later if needed.

For VPS hosting with daily automated backups, you get additional protection at the infrastructure level.

Troubleshooting (with concrete fixes)

Missing config crash loop:
The gateway fails to start and restarts repeatedly.

Fix: Re-run onboarding. This regenerates the configuration

bash
openclaw onboard

Telegram “access not configured”:
Your Telegram bot is set up but OpenClaw can’t access it.

Fix: Approve the pairing:

bash
openclaw pairing approve telegram

Gateway refuses to bind:
The gateway can’t start because the port is already in use or permission is denied.

Fix: Set the gateway mode. If you’re binding to a non-localhost interface, authentication is required.

bash
# Check current mode
openclaw gateway status

# Force restart if locked
openclaw gateway --force

No API key found:
The assistant can’t connect to the LLM.

Fix: Use the onboarding wizard or set the API key as an environment variable:

bash
export OPENAI_API_KEY=your-key-here
# Or for the specific provider you're using

Service fails health check:
The gateway is running but not passing health checks.

Fix: Check the logs to identify the specific issue:

bash
journalctl --user -u openclaw-gateway -f

Or use OpenClaw’s log command:

bash
openclaw logs --follow

Ready to Deploy?

You have the steps. Now you need the right server to run it all.

That’s where we come in. At Truehost, we offer VPS plans built specifically for OpenClaw and AI automation.

Our servers give you dedicated resources, so no other user slows you down.

And, we preload Ubuntu with Playwright, Python, and OpenClaw already installed and ready to go.

Plan	Features	Price
Starter	1 vCPU, 2 GB RAM, 50 GB NVMe, 4 TB bandwidth	₦10,500/month
Pro	2 vCPU, 4 GB RAM, 100 GB NVMe, 6 TB bandwidth	₦22,750/month
Business	4 vCPU, 8 GB RAM, 200 GB NVMe, 8 TB bandwidth	₦56,000/month

You can deploy within seconds. One click from our dashboard provisions your VPS with everything configured. You access it via root SSH, paste your keys, and start your assistant immediately.

We include 24/7 support, Naira payments, DDoS protection, and daily automated backups.