🦞 Building a 24/7 Jarvis: My Journey Setting Up OpenClaw

The AI agent revolution is here, and I finally took the plunge.

After watching OpenClaw go viral across GitHub, I decided it was time to move past basic chatbots and build a true, autonomous personal AI assistant that can actually do things

on my behalf. Here is how my weekend project went, the hardware hurdles I hit, and where I’m taking this next. — ### 💻 The Hardware & The OS Plot Twist I dug out an old, reliable ThinkPad T480s to serve as my dedicated OpenClaw gateway server. Initially, the plan was simple: keep it on Windows 10. However, Windows 10 had actually crashed completely months ago. > The Pivot: Instead of dealing with a re-installation of Windows back then, I wiped the drive and installed Linux Mint. Why Mint? That’s a long story for another day, but honestly, it was the best accidental decision I could have made. It’s lightweight, incredibly stable, and perfect for breathing new life into older hardware. When it came time to set up OpenClaw this weekend, the installation via Node.js was a breeze on it. — ### 🧠 The Brains: OpenRouter vs. Ollama An agent is only as good as its LLM. I wanted to test two distinct paths for the backend: Ollama (Local Inference): I really wanted to keep everything 100% local, so I booted up Ollama. Unfortunately, reality hit hard. Running decent-sized models locally requires serious hardware. With the limited RAM on an older T480s, it quickly became clear that a fully local agent loop was never going to work smoothly here. OpenRouter (Cloud API): I pivoted and set up an account with OpenRouter to handle the API routing to cloud models. Wow. I am seriously impressed. It bridges the gap perfectly—giving my local OpenClaw instance access to flagship, heavy-hitting models with minimal latency and ultra-low cost. — ### 💬 “Why is my AI texting me?” (Setting up Telegram) One of OpenClaw’s coolest features is its multi-channel gateway, allowing you to interface with your agent over apps you already use. I managed to successfully wire it up to Telegram. Right now, if I’m honest, I don’t fully understand the day-to-day use case for this yet. It’s a bit surreal to have a custom bot sitting in my chat list waiting for commands. But having a 24/7 background daemon reachable via text message feels like the foundational plumbing for something

👉 Call to Action: Share your thoughts, try the technique, or follow for more insights.