Use opencode with TurboLLM
opencode is an open-source coding agent that runs in your terminal. Instead of a cloud endpoint, you can point it at a model running on your own GPU — private, offline, and free of per-token billing. TurboLLM does the wiring for you with one command.
One command
From a machine where TurboLLM is running, launch opencode already connected to your local model:
What it does
The launcher writes opencode's own config file the way the tool expects, then starts opencode — so there's no manual copy-paste setup. It points opencode at your local TurboLLM server so every request runs on your GPU.
If you already have an opencode config and TurboLLM can't parse it, it refuses to overwrite the file and instead prints instructions for adding the provider by hand.
Manual setup
If you'd rather wire it up yourself, opencode is OpenAI-compatible, so any tool that speaks the OpenAI API can talk to TurboLLM. Point opencode's provider at your local server:
Base URL: http://localhost:6996/v1
API key: <from Developer → API Keys>
Model: <the model name shown in TurboLLM>The API key is only required if Require API key is turned on in TurboLLM. When it's off, you can leave the key blank or use any placeholder.
The in-app Developer screen shows ready-made snippets for OpenAI-compatible tools, including the exact base URL, key, and model name to paste in.
Recommended models
opencode is a coding agent, so pick a coder-tuned model sized to your GPU. See the model catalog for coder models grouped by GPU size, from small cards to workstation-class hardware.