OpenClaw

OpenClaw speaks Chat Completions through a provider in its config. launch keeps that in a profile of its own, ~/.openclaw-superfluid, so your own OpenClaw (its channels, memory and model) is left as it is.

npm install -g openclaw@latest                     # install
superfluid launch openclaw --model unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_M
superfluid launch openclaw -- agent --local -m "summarize this repo"

Aliases: clawdbot, moltbot.

What launch does

runs	what for
openclaw --profile superfluid setup --baseline	the first time, to create the profile
openclaw --profile superfluid config set models.providers.superfluid ...	a superfluid provider (openai-completions) with the server's /v1, its key and the served model with the server's context window
openclaw --profile superfluid config set agents.defaults.model.primary superfluid/<model>	the served model as the default
openclaw --profile superfluid tui --local	the terminal UI on the embedded runtime, which needs no gateway; arguments after -- replace tui --local