Integrations

Because LLMBoost speaks the OpenAI API, most tools connect by pointing their base URL at LLMBoost. Here are the common ones.

OpenWebUI

A self-hosted chat UI for your models. Connect it to LLMBoost as an OpenAI connection.

In the UI: Settings → Connections → OpenAI API → add http://<llmboost-host>:8000/v1, any API key (e.g. not-needed), save. Your served models appear in the model picker.

Or by env when launching OpenWebUI:

docker run -d -p 3000:8080 \
  -e OPENAI_API_BASE_URL=http://<llmboost-host>:8000/v1 \
  -e OPENAI_API_KEY=not-needed \
  ghcr.io/open-webui/open-webui:main

Agentic coding tools

Point your terminal coding agent at LLMBoost — keep your code and prompts on your own GPUs.

OpenCode

Add LLMBoost as an OpenAI-compatible provider in opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "llmboost": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LLMBoost",
      "options": {"baseURL": "http://localhost:8000/v1"},
      "models": {"deepseek-ai/DeepSeek-V3.2": {}}
    }
  }
}

Then select the llmboost provider/model in OpenCode.

OpenAI Codex

Add a provider in ~/.codex/config.toml (OpenAI-compatible):

model = "deepseek-ai/DeepSeek-V3.2"
model_provider = "llmboost"

[model_providers.llmboost]
name = "LLMBoost"
base_url = "http://localhost:8000/v1"
env_key = "OPENAI_API_KEY"

export OPENAI_API_KEY=not-needed
codex

note

Coding agents work best with a capable, tool-trained instruct model. Serve one sized to your GPUs (see Multi-GPU) and use its full Hugging Face name as the model id. Exact config keys vary by tool version — check each tool's docs for the latest.