OpenClaw’s Two Configuration Methods for Accessing Claude: OpenAI-Compatible Mode vs. Claude Native Format Complete Tutorial

Author's Note: A step-by-step guide to configuring both OpenAI-compatible mode and Claude native format in OpenClaw, including complete JSON configuration code, applicable model lists, and key differences.

In OpenClaw (Open WebUI), there are two ways to connect to Large Language Models: OpenAI-compatible mode (openai-completions) and Claude native format (anthropic-messages). Many users aren't clear on the differences, leading to either using the wrong format for Claude models or missing out on advanced features like Prompt Caching that the native format offers.

Core Value: After reading this article, you'll master the complete configuration methods for both connection types in OpenClaw, know exactly which format to use for each model, and be able to directly copy and use the configuration code.

Core Comparison of Two Connection Methods in OpenClaw

One-Sentence Summary of the Two Connection Methods in OpenClaw

Remember a simple rule: Use anthropic-messages for Claude series models, and openai-completions for all other models. The most obvious difference between the two is the baseUrl—OpenAI compatible mode ends with /v1, while Claude native format doesn't.

OpenClaw OpenAI-Compatible Mode Configuration Tutorial

When to Use OpenAI-Compatible Mode

OpenAI-Compatible Mode (openai-completions) is the most versatile access method in OpenClaw, suitable for all non-Claude Large Language Models. Most API proxy services use this standardized OpenAI format.

Complete Configuration Code for OpenAI-Compatible Mode

Here's the full configuration for accessing GPT-5.4 via APIYI:

{
  "agents": {
    "defaults": {
      "model": { "primary": "apiyi/gpt-5.4" }
    }
  },
  "models": {
    "providers": {
      "apiyi": {
        "baseUrl": "https://api.apiyi.com/v1",
        "apiKey": "sk-your-api-key",
        "api": "openai-completions",
        "models": [
          { "id": "gpt-5.4", "name": "GPT-5.4" }
        ]
      }
    }
  }
}

{
  "models": {
    "providers": {
      "apiyi": {
        "baseUrl": "https://api.apiyi.com/v1",
        "apiKey": "sk-your-api-key",
        "api": "openai-completions",
        "models": [
          { "id": "gpt-5.4", "name": "GPT-5.4" },
          { "id": "gemini-3-flash-preview", "name": "Gemini 3 Flash" },
          { "id": "deepseek-v3.2", "name": "DeepSeek V3.2" },
          { "id": "glm-5", "name": "GLM-5" },
          { "id": "kimi-k2.5", "name": "Kimi K2.5" },
          { "id": "grok-4", "name": "Grok 4" },
          { "id": "Minimax-M2.5", "name": "Minimax M2.5" }
        ]
      }
    }
  }
}

Key Configuration Points for OpenAI-Compatible Mode

🎯 Configuration Reminder: Don't omit the /v1 at the end of the baseUrl—it's the standard path for the OpenAI-compatible protocol. Register at APIYI apiyi.com to get your API Key and free credits.

OpenClaw Claude Native Format Configuration Tutorial

When to Use Claude Native Format

Claude Native Format (anthropic-messages) is the exclusive access method for Claude series models. Using the native format gives you access to Claude's unique advanced features like Prompt Caching, Extended Thinking, and PDF processing.

Complete Configuration Code for Claude Native Format

Here's the full configuration for accessing Claude models via APIYI:

{
  "models": {
    "providers": {
      "apiyi-claude": {
        "baseUrl": "https://api.apiyi.com",
        "apiKey": "sk-your-api-key",
        "api": "anthropic-messages",
        "headers": {
          "anthropic-version": "2023-06-01",
          "anthropic-beta": ""
        },
        "models": [
          {
            "id": "claude-sonnet-4-6",
            "name": "Claude Sonnet 4.6",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          },
          {
            "id": "claude-sonnet-4-6-thinking",
            "name": "Claude Sonnet 4.6 Thinking",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          }
        ]
      }
    }
  }
}

{
  "models": {
    "providers": {
      "apiyi-claude": {
        "baseUrl": "https://api.apiyi.com",
        "apiKey": "sk-your-api-key",
        "api": "anthropic-messages",
        "headers": {
          "anthropic-version": "2023-06-01",
          "anthropic-beta": ""
        },
        "models": [
          {
            "id": "claude-sonnet-4-6",
            "name": "Claude Sonnet 4.6",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          },
          {
            "id": "claude-sonnet-4-6-thinking",
            "name": "Claude Sonnet 4.6 Thinking",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          },
          {
            "id": "claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          },
          {
            "id": "claude-haiku-4-5-20251001",
            "name": "Claude Haiku 4.5",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 8192
          }
        ]
      }
    }
  }
}

Key Configuration Points for Claude Native Format

🎯 Critical Difference: The baseUrl for Claude Native Format does NOT include /v1. This is the most common mistake for beginners—if Claude access fails, first check if you accidentally added /v1 to the URL.

Configuring OpenClaw for Both Formats Simultaneously

In real-world use, you'll likely need to use both general models and Claude models. This requires configuring two providers in OpenClaw:

Dual Provider Combined Configuration Code

Write both format providers in the same configuration file, and you'll be able to switch between models freely in OpenClaw:

{
  "agents": {
    "defaults": {
      "model": { "primary": "apiyi/gpt-5.4" }
    }
  },
  "models": {
    "providers": {
      "apiyi": {
        "baseUrl": "https://api.apiyi.com/v1",
        "apiKey": "sk-your-api-key",
        "api": "openai-completions",
        "models": [
          { "id": "gpt-5.4", "name": "GPT-5.4" },
          { "id": "deepseek-v3.2", "name": "DeepSeek V3.2" },
          { "id": "gemini-3-flash-preview", "name": "Gemini 3 Flash" },
          { "id": "glm-5", "name": "GLM-5" },
          { "id": "kimi-k2.5", "name": "Kimi K2.5" },
          { "id": "grok-4", "name": "Grok 4" },
          { "id": "Minimax-M2.5", "name": "Minimax M2.5" }
        ]
      },
      "apiyi-claude": {
        "baseUrl": "https://api.apiyi.com",
        "apiKey": "sk-your-api-key",
        "api": "anthropic-messages",
        "headers": {
          "anthropic-version": "2023-06-01",
          "anthropic-beta": ""
        },
        "models": [
          {
            "id": "claude-sonnet-4-6",
            "name": "Claude Sonnet 4.6",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          },
          {
            "id": "claude-sonnet-4-6-thinking",
            "name": "Claude Sonnet 4.6 Thinking",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          },
          {
            "id": "claude-opus-4-6",
            "name": "Claude Opus 4.6",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 200000,
            "maxTokens": 16384
          }
        ]
      }
    }
  }
}

🎯 Important Note: Both providers can use the same API Key. A single key from APIYI apiyi.com supports both the OpenAI-compatible format and the native Claude format—you don't need to apply for multiple keys.

Common Troubleshooting for Both Formats in OpenClaw

The most common mistakes during configuration involve mismatched baseUrl and api types. Here are the common errors and their solutions:

🎯 Quick Troubleshooting Tip: OpenAI format includes /v1, Claude format doesn't. Remember this and you'll avoid 80% of configuration errors. If you encounter other issues, you can visit the APIYI apiyi.com documentation center for the complete integration guide.

Frequently Asked Questions

Summary

Here are the key points for the two OpenClaw integration methods:

1. Use openai-completions for General Models: GPT, Gemini, DeepSeek, GLM, Kimi, Grok, Minimax, and all other non-Claude models. The baseUrl should include /v1.
2. Use anthropic-messages for Claude Models: claude-sonnet-4-6, claude-opus-4-6, claude-haiku, etc. The baseUrl should NOT include /v1, and you need the anthropic-version header.
3. Having both Providers is the best practice: Configure two providers with the same API Key to freely switch between all models within OpenClaw.

We recommend getting your API Key from APIYI (apiyi.com). A single key gives you access to all major models like GPT, Claude, Gemini, and DeepSeek, supporting both OpenAI-compatible and Claude native formats.

📚 References

1. APIYI Help Center: Complete OpenClaw Integration Guide

   • Link: help.apiyi.com
   • Description: Contains detailed integration documentation and the latest model lists for all sites

2. Anthropic API Documentation: Claude Native API Format Specifications

   • Link: platform.claude.com/docs/en/api/messages
   • Description: Complete parameters and response formats for the Messages API

3. OpenAI SDK Compatibility Documentation: Which parameters are ignored on Claude

   • Link: platform.claude.com/docs/en/api/openai-sdk
   • Description: Full list of supported and unsupported parameters

4. Open WebUI Documentation: OpenClaw Multi-Provider Configuration Guide

   • Link: docs.openwebui.com
   • Description: Provider configuration, model management, and Agent settings

Author: APIYI Technical Team
Technical Discussion: Feel free to discuss in the comments. For more resources, visit the APIYI documentation center at docs.apiyi.com