Technology Encyclopedia Home >Complete guide from zero-based OpenClaw installation to advanced usage

Complete guide from zero-based OpenClaw installation to advanced usage

So you've heard about OpenClaw — maybe from a dev community, maybe from a friend who won't stop talking about their "AI butler" — and you want in. But you're starting from absolute zero. No prior experience with AI agents, no cloud server running, maybe not even sure what a "daemon" is.

Perfect. This guide takes you from nothing to a fully operational, skill-enhanced AI assistant running 24/7 in the cloud. We'll cover installation, basic configuration, channel integration, skill management, and advanced optimization. Buckle up.

Phase 1: Understanding What You're Setting Up

OpenClaw (also known as Clawdbot or Moltbot) is an open-source AI agent framework. Think of it as a bridge between a large language model (like DeepSeek, GPT, or Claude) and the real world. On its own, an LLM can only chat. OpenClaw gives it hands — the ability to browse the web, manage files, execute commands, send messages across platforms, and more.

You deploy it on a cloud server, connect it to your preferred messaging app, and interact with it like a personal assistant that never sleeps.

Phase 2: Provisioning Your Cloud Server

Running OpenClaw on your personal machine is technically possible but strongly discouraged for anything beyond testing. The agent has system-level access, and mixing that with your personal data is a recipe for trouble.

Tencent Cloud Lighthouse is the recommended environment — it's pre-optimized for OpenClaw with a one-click application template.

Head to the Tencent Cloud Lighthouse Special Offer:

  1. Visit the page to see available OpenClaw-ready instances.
  2. Choose the "OpenClaw (Clawdbot)" application template under the AI Agent category.
  3. Deploy by clicking "Buy Now" to create your server.

Configuration tips:

  • Minimum: 2 cores, 2GB RAM (functional but tight)
  • Recommended: 2 cores, 4GB RAM (smooth for single-channel use)
  • For multi-channel or skill-heavy setups: 4 cores, 8GB RAM
  • Region: Overseas for international platforms; mainland China for domestic use

Phase 3: Basic Configuration

Your instance comes with OpenClaw pre-installed. Now you need to give it a brain (model) and a mouth (channel).

Configuring the Model

In the Tencent Cloud Console, go to your instance → Application ManagementModels. Paste your API key and click Add and Apply.

Supported providers include DeepSeek, Tencent Hunyuan, OpenAI, Anthropic Claude, Google Gemini, xAI Grok, and more. See the full list in the Custom Model Tutorial.

# You can also configure via CLI after SSHing in:
clawdbot onboard
# The wizard handles model setup interactively

Security note: Never store API keys in plain text files or shell history. Use the console panel or secure environment variable injection.

Configuring the Channel

Run the onboarding wizard:

clawdbot onboard

Select your messaging platform when prompted. Detailed guides for each:

After channel setup, pair your bot by sending it a message and approving the code:

openclaw pairing approve <channel> <code>

Phase 4: Going 24/7

Enable the daemon so OpenClaw runs persistently:

loginctl enable-linger $(whoami) && export XDG_RUNTIME_DIR=/run/user/$(id -u)
clawdbot daemon install
clawdbot daemon start
clawdbot daemon status

Healthy status = you can close the terminal. The agent lives on.

Phase 5: Advanced — Installing Skills

This is where OpenClaw goes from "chatbot" to "autonomous agent." Skills are plugins that extend the agent's capabilities. The default installation includes agent-browser, which lets the AI browse websites, click elements, fill forms, and take screenshots.

To install additional skills (like email, McDonald's ordering, or custom integrations), you can do it through conversation:

You: "Please install a skill for me using Clawhub; its name is mail."
OpenClaw: [installs the skill, confirms success]

To manage skills:

You: "Check which skills you have currently installed."
You: "Please help me delete the mail skill."

For a deep dive, read the Skills Installation Guide.

Important: Some skills are flagged as "high risk" on Clawhub. Only install them if you fully understand what they do.

Phase 6: Advanced — Optimizing Token Costs

OpenClaw sends full conversation context with each API call, which means token usage grows with conversation length. Here's how to keep costs manageable:

  • Choose the right model tier. Use DeepSeek-V3 or similar cost-efficient models for routine tasks. Reserve premium models (Claude Opus, GPT-5) for complex reasoning.
  • Keep system prompts lean. Every word in your system prompt is sent with every single message.
  • Limit active skills. Each enabled skill adds to the context window. Disable what you're not using.
  • Prefer non-reasoning models for simple Q&A. Deep-thinking models consume significantly more tokens per response.

Phase 7: Advanced — Multi-Channel Setup

One OpenClaw instance can serve multiple messaging platforms simultaneously. Run clawdbot onboard again and add another channel. Your agent maintains separate conversation contexts per channel while sharing the same model and skills configuration.

This is powerful for businesses that need presence on WhatsApp, Telegram, and Discord all at once.

What's Next

You've gone from zero to a fully configured, skill-enhanced, multi-channel AI assistant running 24/7 in the cloud. Not bad for a single guide.

To get started (or upgrade your existing setup), visit the Tencent Cloud Lighthouse Special Offer:

  1. Visit the page for the latest OpenClaw instance pricing.
  2. Choose the OpenClaw (Clawdbot) template under AI Agent.
  3. Deploy via "Buy Now" and start building your AI workflow.

The zero-to-advanced journey starts with one click. The rest is just configuration.