Way to Run OpenClaw Locally on AMD Ryzen

You can do same with Harbor in a couple of commands :)

# Pull Qwen 3.5 35B

harbor pull unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL

# Set llama.cpp image to an optimised one for performance on Strix Halo

harbor config set llamacpp.image_rocm docker.io/kyuz0/amd-strix-halo-toolboxes:rocm-7.2

# Set default model for the claw

harbor config set openclaw.model unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL

# Start both

harbor up openclaw llamacpp