OpenClaw on Strix Halo: Self-Hosted AI Agent Setup (2026)

Set up OpenClaw on AMD Strix Halo with 128GB unified memory — from fresh Ubuntu Server to a working self-hosted AI agent. Kernel 6.18, ROCm 7.2, 100GB GPU memory, Ollama with GLM-4.7-flash, and the full interactive onboarding walkthrough. No cloud, no API keys, everything local. Written guide with all commands: https://www.localhake.com/content/openclaw-strix-halo-setup Recommended Gear Some links are affiliate links — I may earn a commission at no extra cost to you. GMKtec EVO X2 (AI MAX+ 395, 128GB) - Featured in this video: https://amzn.to/4rJsga0 Minisforum MS-S1 MAX (AI MAX+ 395, 128GB): https://amzn.to/4uUZMg7 Beelink GTR9 Pro (AI MAX+ 395, 128GB): https://amzn.to/3PO5gt4 GEEKOM A9 Mega (AI MAX+ 395, 128GB): https://amzn.to/4lWatuZ Chapters 0:00 Intro 0:30 Prerequisites 3:50 Terminal Setup 4:20 System Update & Baseline 5:40 Kernel 6.18 Download 7:00 Kernel Install & Reboot 8:40 ROCm 7.2 Install 10:40 Unified Memory (TTM) 12:30 HSA Environment Variables 13:40 GPU Verification 14:50 Ollama Install 16:00 Ollama Strix Halo Config 17:40 Node.js & OpenClaw 19:00 Interactive Onboarding 24:40 SSH Tunnel & Dashboard 26:40 First Agent Task 28:40 Service Check Demo 30:20 Reboot Test 31:40 What's Next This is Part 1 of a 6-part series: Part 1: Agent Setup (this video) Part 2: AI Incident Response with Uptime Kuma https://youtube.com/watch?v=o9t7_1Mn8iE Part 3: Self-Hosted Matrix Command Center https://youtube.com/watch?v=MfD00eDPnxU Part 4: Local Model Shootout https://youtube.com/watch?v=7KOebtGHWm0 Part 5: Backup & Restore https://youtube.com/watch?v=2cpjPzEVscU Part 6: HTTPS & Clean URLs with Caddy https://youtube.com/watch?v=69UZ7HApkgI #openclaw #strixhalo #homelab #selfhosted #localai