Use openclaw with TurboLLM
openclaw is a coding agent. With TurboLLM you can point it at a model running on your own GPU — one command wires it up and launches it, fully offline.
One-command launch
Run this and TurboLLM wires openclaw to your local server, then launches it:
TurboLLM edits openclaw's config to add the local provider and then starts openclaw for you. If the config can't be parsed, TurboLLM refuses to overwrite it and prints exactly how to add the provider by hand — it will never clobber a config it doesn't understand.
openclaw's launcher config shape is best-effort — its format can change between versions. If the first run doesn't connect, don't worry: use the manual base-URL method below. It always works.
Manual fallback
openclaw speaks the OpenAI-compatible API, so you can wire it up by hand in its provider settings. Point it at TurboLLM's local endpoint:
Base URL: http://localhost:6996/v1
API key: any non-empty valueThe key isn't checked — TurboLLM runs locally on your machine — but most clients require some value in the field, so put any placeholder there.
Recommended models
Coding agents like openclaw work best with capable instruction- and tool-following models. See the model catalog for options that fit your GPU.