I’ve configured OpenClaw to use OpenAI as its primary model and automatically fall back to a locally hosted Ollama model if cloud access or usage limits become an issue.

Here’s the basic setup:

1. Install Ollama

Download it from:
https://ollama.com/download

2. Download a model

ollama pull gemma4:e4b

You can check installed models with:

ollama list

3. Let OpenClaw discover Ollama

export OLLAMA_API_KEY="***"
openclaw models list --provider ollama

For an entirely local setup:

openclaw models set ollama/gemma4:e4b

Alternatively, keep a cloud model as the primary and add Ollama as the fallback:

{
  agents: {
    defaults: {
      model: {
        primary: "openai/gpt-5.6-sol",
        fallbacks: ["ollama/gemma4:e4b"]
      }
    }
  }
}

The result? Cloud performance when available, with a private local backup ready to take over when needed—no frantic shouting at a usage-limit screen.

Always use the exact model name shown by ollama list.

Official guide:
https://docs.openclaw.ai/providers/ollama/setup

#OpenClaw #Ollama #LocalAI #SelfHosted #ArtificialIntelligence #OpenSourceAI