Model Provider Quickstart: Connect and Configure LLMs
Learn how to connect a model provider to OpenClaw, complete authentication, and set your default model using the provider/model format. This page suits anyone configuring an LLM for the first time.
Read this when
- You want to choose a model provider
- You want quick setup examples for LLM auth + model selection
Select a provider, complete authentication, then assign the default model using provider/model.
Quick start (two steps)
- Log in to your chosen provider (typically through
openclaw onboard). - Define the default model:
{
agents: { defaults: { model: { primary: "anthropic/claude-opus-4-6" } } },
}
Supported providers (starter set)
- Alibaba Model Studio
- Amazon Bedrock
- Anthropic (API + Claude CLI)
- Baseten (Inkling + Model APIs)
- BytePlus (International)
- Chutes
- Cloudflare AI Gateway
- Cohere
- ComfyUI
- DeepInfra
- fal
- Fireworks
- MiniMax
- Mistral
- Moonshot AI (Kimi + Kimi Coding)
- NovitaAI
- OpenAI (API + Codex)
- OpenCode (Zen + Go)
- OpenRouter
- Qianfan
- Qwen
- Runway
- StepFun
- Synthetic
- Venice (Venice AI)
- Vercel AI Gateway
- xAI
- Z.AI (GLM)
The complete list of providers plus settings for advanced use cases lives in Provider directory and Model providers.
Additional provider variants
anthropic-vertex- install@openclaw/anthropic-vertex-providerto get implicit Anthropic support on Google Vertex. That capability kicks in whenever Vertex credentials exist. No dedicated auth option appears during onboarding.copilot-proxy- a local bridge to the VS Code Copilot Proxy. Useopenclaw onboard --auth-choice copilot-proxy.google-gemini-cli- an optional explicit runtime for canonicalgoogle/*models. A localgeminiinstall is needed, along with a supported Google AI Studio API-key profile. Setting up Gemini CLI or Antigravity OAuth from scratch is not available.