One-click deployment of OpenClaw on VPS: Select China-optimized line, plus a method to get a free 450 RMB API quota | Warner Cloud Experience

This article isOverseas Network Special ReportIssue 12

This article was generated by Gemini based on my video script.

Foreword: Creating Your Own AI Assistant

I've been doing this lately. OpenClaw(An open-source AI agent that can execute code on your terminal or system). For ease of use anytime, anywhere, I chose to use... VPS + 1 Panel The combination of these can be deployed in the cloud.

This article will share the detailed deployment process, cost-saving tips for large model APIs, and my actual test of the Warner Cloud server lines.

Chapter 1: Server Environment Preparation (Warner Cloud Real-World Test)

I tested two WarnerCloud lightweight cloud servers:

Registration link:WarnerCloud

• Hong Kong, China (1 core, 1GB): Suitable for China Mobile broadband users, the return trip is basically... CMIN2 Top-tier lines with extremely low latency (approximately 10ms).
• Tokyo, Japan (2 cores, 4GB RAM): China Telecom, China Unicom, and China Mobile usually go through... AS4837 With optimized lines and ample bandwidth, latency across all three major networks is between 50-70ms, offering exceptional value for money.

Note: If your IP address is "sent to China" and this affects your AI service unlocking, you can resolve this by changing your outbound IP address or configuring a WARP.

【IP address forwarding troubleshooting tutorial: See here】

Evaluation data:

- Hongkong:https://paste.spiritlhl.net/#/show/xM2Gs.txt

Tokyo:https://paste.spiritlhl.net/#/show/Ax9sk.txt

Chapter 2: One-Click Deployment of OpenClaw using 1Panel

Deploying via 1Panel is very simple:

1. Install 1Panel: Execute the installation code from the official website in the VPS terminal.
2. Creating intelligent agents: In 1Panel, click Create in the "AI - Intelligent Agent" menu.
3. Configuration model: This is a crucial step; you need to provide the API Key and URL for the large language model.

Chapter 3: How to "get" the large model API for free?

Since OpenAI's API is relatively expensive, I recommend using a domestic API aggregation platform. I use... Baishan Intelligence.

• wool: Personal verification grants a 150 RMB credit limit, with an additional 300 RMB bonus on the first use (limited-time offer), totaling... 450 yuan.
• Supported models: DeepSeek V3.2, Qianwen 30B, MiniMax, Zhipu GLM-4, etc.
• Configuration method: Generate an API Key using Baishan Smart Computing and enter it into the "Custom Model Provider" settings in 1Panel.

Chapter 4: Lark Integration and Actual Costs

For ease of use, I have integrated OpenClaw.FeishuYou can chat with AI directly on your phone and have it perform tasks.

• Experience: The model may sometimes "slack off" or get stuck in an infinite loop of repeated output, which requires manual correction (it is recommended to try different models, such as DeepSeek or Qianwen Coder).
• Actual cost measurement: Using the DeepSeek V3.2 model, asking a simple question takes approximately [time/percentage]. 0.02 yuanHowever, if the model gets stuck in an infinite loop and keeps repeating the same output, solving a complex problem might cost 2-3 yuan. Overall, the free credit can be used for a long time.

Summarize

Deploying OpenClaw via VPS significantly reduces local hardware requirements. Combined with free API quotas, you can experience this powerful AI agent at zero cost. In the future, I will also share how to remotely call large models on your local Mac using Tailscale, so stay tuned.