iTranslated by AI

The content below is an AI-generated translation. This is an experimental feature, and may contain errors. View original article
🦞

I tried OpenClaw and it was more incredible than I imagined

に公開

About This Article

I installed an OSS autonomous AI agent called OpenClaw on my VPS, so I'm sharing the setup process and my impressions of using it.

What is OpenClaw?

OpenClaw is an MIT-licensed self-hosted AI agent. It has over 200,000 stars on GitHub and functions as a gateway integrating chat platforms like WhatsApp, Telegram, Discord, and Slack.

In a nutshell, it's an "AI that lives on your server and works 24/7."

How is it different from ChatGPT?

ChatGPT OpenClaw
Operating Environment OpenAI's servers Your own VPS
Activation Timing Only when you open the browser 24/7 resident
Proactive Actions ❌ Cannot ✅ Scheduled via cron
Command Execution ❌ Cannot ✅ Shell operations possible
Chat Integration ChatGPT app only Discord, Slack, WhatsApp, etc.

The biggest difference is "whether it can act on its own." ChatGPT does nothing unless you talk to it, but OpenClaw, if scheduled with cron for regular tasks, can summarize news every morning or detect server anomalies and send alerts.

Set up

Environment

  • VPS: XServer VPS (2GB RAM)
  • OS: Ubuntu 24.04
  • Node.js: v22
  • AI Model: OpenAI GPT-5.3-Codex (ChatGPT Plus OAuth authentication)

Steps

# Install
npm install -g openclaw@latest

# Initial setup wizard
openclaw onboard --install-daemon

# Start gateway
openclaw gateway --port 18789

By simply following the wizard, you can complete everything from AI model authentication to channel integration. I'm operating it as a Discord Bot, so I created a bot in the Discord Developer Portal and set up the token.

Daemonize with systemd

I registered it as a systemd service so that it automatically starts even after a VPS reboot.

# /etc/systemd/system/openclaw.service
[Unit]
Description=OpenClaw Gateway
After=network.target

[Service]
Type=simple
User=openclaw
Environment=NVM_DIR=/home/openclaw/.nvm
ExecStart=/bin/bash -c 'source /home/openclaw/.nvm/nvm.sh && openclaw gateway --port 18789'
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target
systemctl enable openclaw
systemctl start openclaw

Now the Gateway will launch automatically just by powering on.

Impressions After Using It

The Experience of AI Living in Discord

Being able to converse with AI directly in the Discord server channel you normally use was more comfortable than I imagined. The hassle of opening a browser and navigating to ChatGPT becomes zero.

exec is amazing

What personally impressed me the most was the exec tool. If I send "check disk space" in Discord, it actually executes df -h on the VPS and returns the result. Being able to manage the server without SSH is revolutionary.

Customizing Personality

You can freely change the agent's name and tone of voice. I set it to be a bit sarcastic, and it's quite enjoyable.

Value for Individual Developers

I develop web applications personally, and since integrating OpenClaw, my efficiency has subtly increased.

  • Server checks can be done via chat
  • Routine health checks can be automated
  • Research can be casually offloaded
  • Drafts for technical articles can be generated

In particular, being able to automate "tasks I intend to do when I have free time but then forget" with cron is a huge benefit.

Points to Note

  • VPS specifications: minimum 2GB memory recommended
  • Security settings should be properly configured (e.g., allowlist)
  • AI model API fees are separate (no additional cost with ChatGPT Plus OAuth)

Conclusion

Honestly, I started using it with a feeling of "Oh, another AI tool," but after trying it, it was completely different from ChatGPT. The phrase "not a tool to converse with AI, but an infrastructure where AI works for you" feels just right.

It is especially recommended for individuals who own a VPS and are engaged in personal development.

https://github.com/nicholasgasior/openclaw

GitHubで編集を提案

Discussion