OpenClaw Setup

OpenClaw is an open-source AI agent platform that can use model-go as a model provider.

What is OpenClaw?

OpenClaw is a local AI assistant framework that supports:

• Multi-channel messaging such as Telegram, Discord, and WhatsApp
• Tool calling such as browser, filesystem, and shell access
• Memory systems and skill extensions

Install OpenClaw

npm install -g openclaw

Configure model-go as a provider

Edit ~/.openclaw/openclaw.json and add model-go under models.providers.

Anthropic models (Claude)

{
  "models": {
    "providers": {
      "model-go": {
        "baseUrl": "https://model-go.com",
        "apiKey": "your model-go token",
        "auth": "api-key",
        "api": "anthropic-messages",
        "authHeader": true,
        "models": [
          {
            "id": "replace with a Claude model name from the Models page",
            "name": "replace with a Claude model name from the Models page",
            "api": "anthropic-messages",
            "reasoning": true,
            "input": ["text", "image"],
            "contextWindow": 200000,
            "maxTokens": 8192,
            "compat": {
              "supportsReasoningEffort": true,
              "maxTokensField": "max_tokens"
            }
          }
        ]
      }
    }
  }
}

OpenAI models (GPT / Codex)

{
  "models": {
    "providers": {
      "model-go-openai": {
        "baseUrl": "https://model-go.com/v1",
        "apiKey": "your model-go token",
        "api": "openai-responses",
        "models": [
          {
            "id": "replace with an OpenAI model name from the Models page",
            "name": "replace with an OpenAI model name from the Models page",
            "api": "openai-responses",
            "reasoning": true,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          }
        ]
      }
    }
  }
}

Google models (Gemini)

{
  "models": {
    "providers": {
      "model-go-gemini": {
        "baseUrl": "https://model-go.com/v1beta",
        "apiKey": "your model-go token",
        "auth": "api-key",
        "api": "google-generative-ai",
        "models": [
          {
            "id": "replace with a Gemini model name from the Models page",
            "name": "replace with a Gemini model name from the Models page",
            "api": "google-generative-ai",
            "reasoning": true,
            "input": ["text", "image"],
            "contextWindow": 1000000,
            "maxTokens": 65536
          }
        ]
      }
    }
  }
}

Configuration Notes

Field	Purpose
baseUrl	Anthropic models use https://model-go.com; OpenAI models use https://model-go.com/v1; Gemini models use https://model-go.com/v1beta
apiKey	The API token you created in model-go
api	API protocol: anthropic-messages, openai-responses, or google-generative-ai
auth	Authentication mode, set to api-key
authHeader	Required for Anthropic models
models	Model list with ids, names, and capabilities

Using models

After configuration, you can reference models by provider and model name:

model-go/replace with a Claude model name from the Models page
model-go-openai/replace with an OpenAI model name from the Models page
model-go-gemini/replace with a Gemini model name from the Models page

Set a default model in openclaw.json:

{
  "models": {
    "default": "model-go/replace with a Claude model name from the Models page"
  }
}

Apply the configuration

After editing openclaw.json, restart the gateway:

openclaw gateway restart

TIP

OpenClaw supports model switching at runtime. After configuring multiple providers, you can switch models with the /model command in chat.

Troubleshooting

Model switching fails

If OpenClaw does not recognize the new model, the model cache may be stale.

rm ~/.openclaw/agents/main/agent/models.json
openclaw gateway restart

Connection fails

• Confirm the baseUrl is correct
• Confirm the apiKey is valid and the token group includes the target model
• Run openclaw gateway restart to reload the configuration