πŸ€– OpenClaw Support Training Portal

A simplified, clear framework for managing our #1 deployed Docker platform.

1. Architecture & The Analogy

OpenClaw is an open-source platform hosting a 24/7 personal AI agent that routes complex prompts from chat apps out to live artificial intelligence models.

🏠 The House Analogy:
  • VPS Server = The physical house (always powered on).
  • Docker Environment = A dedicated room inside the house.
  • OpenClaw Software = The physical assistant living in that room.
  • AI Model (LLM) = The assistant's brain (Claude, GPT-4, etc.).
  • Messaging Channels = The phone lines used to text the assistant (WhatsApp/Telegram).

2. Setup Types & Identification

Always ask a client: "Are you using Docker Manager or Standalone?" This single question defines your troubleshooting boundary.

Installation Path Where It's Managed Support Boundaries Status
Docker Manager hPanel GUI (Visual One-Click) Full Infrastructure Support Recommended (90% of Users)
Standalone Template Terminal / SSH Manual Script Limited Support (Redirect to Docs) Deprecated / Dead

How to verify:

Go to hPanel βž” VPS βž” Docker Manager βž” Projects. If OpenClaw isn't explicitly listed here, it is a Standalone deployment.

πŸ’¬ Script for Standalone Inquiries:
"We provide limited support for standalone installations. For app-level troubleshooting, please use the official guide at https://docs.openclaw.ai/troubleshooting. To execute commands, access the container shell via docker exec -it <container_name> bash. I highly recommend migrating to our Docker Manager template for full automated support."

3. Deployment Flows

Path A: Brand New Checkout

The client builds the VPS with OpenClaw preselected during payment setup. A mandatory initialization window will populate the default settings.

Path B: Existing Server Injection

User navigates to the Docker Catalog inside hPanel and triggers deployment.
Warning: If their active OS isn't 'Ubuntu with Docker', changing it will permanently wipe all current server data. Backups are required.

4. The Top-Down Troubleshooting Pipeline

A green "Running" badge inside hPanel only means the container is powered onβ€”it does not mean the app is running smoothly internally. Track down faults using this strict hierarchy:

VPS Server Running? βž” Docker Container Active? βž” Gateway Token Matching? βž” AI Credits Active? βž” Chat App Linked?

Check 1: The Chat Test ⭐

The Ultimate Isolation Shortcut. Have the client click the Chat tab inside the dashboard and type "Hello".

  • Bot Answers: Hosting, keys, and base software are 100% fine. The bug is exclusively a platform pairing error (WhatsApp/Telegram).
  • Bot stays dead: The underlying platform infrastructure layer is broken.

Check 2: Instances Tab

Verifies communication sockets. If a platform badge reads Unreachable or Disconnected, click Manage βž” Restart in hPanel to force a clean socket reload. If it remains broken, it must be re-paired.

Check 3: Profile Setup

Ensure the agent profile (typically named main) has an active model selected under its Model Selection panel. Running blank means it has no active brain.

🚨 Blindspot Warning: Bots Lie About Themselves
OpenClaw agents do not reliably know their own system variables. A bot can text you through WhatsApp claiming its WhatsApp connection is down. Never diagnose problems based on what the bot tells you in textβ€”always check the real indicators on the Instances page or run a direct key test.

5. The Gateway Token (Dashboard Password)

This is the #1 customer issue. If a user loses or skips saving their password token during the initial build window, they are locked out. You never need to reset it. It is permanently stored in clear text.

How to fetch it:

Go to Docker Manager βž” Projects βž” OpenClaw βž” Manage βž” Environment Variables. Locate the value associated with OPENCLAW_GATEWAY_TOKEN.

6. AI Models & Connectivity Diagnostics

OpenClaw can run on unified Nexos Credits (managed through Hostinger) or the user's Personal API Keys (OpenAI, Anthropic).

To verify if a token is dead or alive, run this models request from the VPS command terminal:

curl -L --url 'https://api.nexos.ai/v1/models' --header 'Authorization: Bearer INSERT_NEXOS_KEY_HERE'

Reading Server Return Codes:

βœ… What You CAN Support

  • Deploying OpenClaw through Docker Manager.
  • Retrieving the Gateway Token from settings.
  • Verifying and restarting unresponsive containers.
  • Explaining basic Nexos credit billing concepts.

❌ What You CANNOT Support

  • Agent behavioral problems or low text quality.
  • Custom skills, webhooks, or execution nodes.
  • Scanning WhatsApp QR codes or Telegram BotFather setup.
  • Teaching users prompt engineering or agent logic.