Integrate LiteLLM Models in Openclaw

Table of contents: [Show]

Quick Start

LiteLLM is a unified proxy that lets you access multiple LLM providers through a single OpenAI-compatible API. This guide shows you how to integrate LiteLLM with Openclaw to route requests to Claude, GPT, and other models seamlessly.

Step 1: Start the LiteLLM Proxy

First, ensure your LiteLLM proxy is running locally or remotely. By default, LiteLLM runs on http://localhost:4000.

pip install 'litellm[proxy]'
litellm --model claude-opus-4-6

Step 2: Configure Openclaw

Method A: Interactive Setup (Recommended)

You can quickly configure LiteLLM using Openclaw's onboarding wizard:

Openclaw onboard --auth-choice litellm-api-key

Method B: Manual Configuration

If you prefer to define the exact models available through your proxy, you can edit your ~/.Openclaw/Openclaw.json configuration file. You must set the api format to "openai-completions" and provide the model definitions.

{
  "models": {
    "providers": {
      "litellm": {
        "baseUrl": "http://localhost:4000",
        "apiKey": "${LITELLM_API_KEY}",
        "api": "openai-completions",
        "models": [
          {
            "id": "claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "reasoning": true,
            "input": ["text", "image"],
            "contextWindow": 200000,
            "maxTokens": 64000
          },
          {
            "id": "gpt-4o",
            "name": "GPT-4o",
            "reasoning": false,
            "input": ["text", "image"],
            "contextWindow": 128000,
            "maxTokens": 8192
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": { "primary": "litellm/claude-opus-4-6" }
    }
  }
}

Step 3: Set Your Primary Model

Once configured, set your preferred model via CLI:

Openclaw models set litellm/claude-opus-4-6

Testing Your Setup

Send a test message to verify the integration:

{ "message": "Hello from LiteLLM via Openclaw!" }

Troubleshooting & Best Practices

Connection refused: Ensure LiteLLM proxy is running on the configured port.
Authentication errors: Verify your LITELLM_API_KEY is correctly set.
Model not found: Check that the model ID in your config matches the LiteLLM proxy configuration.
Performance: For production, consider running LiteLLM on a dedicated server rather than localhost.

For more Openclaw integrations, check out vLLM Models Openclaw and GLM Models Openclaw.