Descriptions:
This video from David Ondrej goes beyond basic OpenClaw installation to show how to build the most capable possible OpenClaw configuration on a virtual private server. Starting from a Hostinger one-click OpenClaw deployment, he covers API key setup for Anthropic, model selection between Claude Opus 4.6 and Sonnet 4.5, and a series of advanced customizations that turn the default agent into a self-managing system.
The standout technique is using OpenClaw to reconfigure itself: Ondrej instructs the agent to search the web for the official Anthropic model IDs for Opus and Sonnet, save the results as a persistent markdown reference file, then update its own configuration to add /opus and /sonnet shortcut commands — and finally restart its own gateway to apply the changes. This loop of self-modification, web research, and knowledge persistence illustrates a broader philosophy: rather than repeating deep-research queries in ChatGPT or Perplexity where results are lost, users should direct queries through an agent running on their own VPS and save outputs as permanent markdown files.
The video also details how to recover a gateway token from the terminal if it’s lost, how to verify which model is actively being used by inspecting conversation logs, and how the Mac Mini versus VPS cost tradeoff plays out in practice. It’s aimed at users who already have OpenClaw running but want to move from a baseline setup to a production-grade configuration.
📺 Source: David Ondrej · Published February 09, 2026
🏷️ Format: Tutorial Demo
