Multi-AI MCP for Claude Code

While you can ask for 'all AIs' to get a comprehensive view, you can also target specific models based on their perceived strengths:

• Gemini: Best for technical deep-dives, algorithm complexity analysis, database design, and detailed technical explanations.
• Grok: Best for creative solutions, unconventional approaches, brainstorming, and unique perspectives.
• ChatGPT: Best for balanced analysis, practical advice, and providing code examples for patterns.
• DeepSeek: Best for mathematical problems, advanced coding challenges, and optimizing algorithm performance.

Once installed, you don't need to remember specific commands. You can simply ask Claude Code to perform tasks using the integrated AI tools. The server provides several specialized tools that Claude will invoke based on your request:

• Get multiple opinions: Ask Claude to use ask_all_ais to get perspectives from all enabled AIs.
• Debug code: Ask Claude to use grok_debug or other debug tools for specific models.
• Debate topics: Ask Claude to use ai_debate to have two AIs (e.g., Gemini and ChatGPT) argue a technical decision.
• Code reviews: Ask for reviews using tools like gemini_code_review, openai_code_review, or grok_code_review with specific focuses (e.g., 'technical accuracy' or 'best practices').
• Brainstorming: Use tools like grok_brainstorm for creative input.
• Architecture: Use gemini_architecture for design recommendations.

You can manage your AI providers, specific models, and enable/disable them by editing the credentials.json file located at ~/.claude-mcp-servers/multi-ai-collab/credentials.json.

Supported keys per provider:

• api_key: Your provider's API key.
• model: The specific model string to use (e.g., gemini-2.0-flash, grok-3, gpt-4o).
• enabled: A boolean to turn the provider on or off.

{
  "gemini": {
    "api_key": "your-gemini-key",
    "model": "gemini-2.0-flash",
    "enabled": true
  },
  "grok": {
    "api_key": "your-grok-key",
    "model": "grok-3",
    "enabled": true
  },
  "openai": {
    "api_key": "your-openai-key",
    "model": "gpt-4o",
    "enabled": true
  }
}

If you encounter issues with the Multi-AI MCP, use the following steps:

Verify MCP visibility: Run claude mcp list. You should see multi-ai-collab in the list.

Check server status: If an AI is not responding, run mcp__multi-ai-collab__server_status to check which AIs are enabled and working.

Fix connection errors:

• Verify your API keys are correct in ~/.claude-mcp-servers/multi-ai-collab/credentials.json.
• Ensure you have sufficient API credits for the respective services.

Perform a clean reinstall: If the installation is corrupted, remove it and run the setup again:

claude mcp remove multi-ai-collab
./setup.sh

You can use the multi-AI setup for several common developer tasks:

• Debugging: Ask all AIs for thoughts on a performance issue, or target a specific AI (like Grok) to debug a failing function.
• Code Reviews: Ask all AIs to review a function for security issues, or ask a specific AI (like Gemini) to review a database query for performance bottlenecks.
• Learning: Ask all AIs to explain a new technology (like Docker) to see different pedagogical styles (technical, creative, balanced, or mathematical).
• Brainstorming: Ask Grok specifically to brainstorm creative or 'wild' features for an application.

To get multiple perspectives on a single prompt, use the ask_all_ais tool. This will trigger a request to every currently configured and enabled AI model and return a formatted report containing all responses.

Arguments:

• prompt (string, required): The question or instruction.
• temperature (number, optional): Controls randomness (0.0-1.0, default: 0.7).

{
  "method": "tools/call",
  "params": {
    "name": "ask_all_ais",
    "arguments": {
      "prompt": "What are the pros and cons of using microservices?",
      "temperature": 0.7
    }
  }
}

The ai_debate tool facilitates a structured debate between two AI models. One AI presents an opening argument in favor of a topic, and the second presents a counter-argument.

Arguments:

• topic (string, required): The subject of the debate.
• ai1 (string, optional): The first AI (e.g., gemini, grok, openai, deepseek). Defaults to gemini.
• ai2 (string, optional): The second AI. Defaults to grok.

{
  "method": "tools/call",
  "params": {
    "name": "ai_debate",
    "arguments": {
      "topic": "Is AI-driven coding a threat to junior developers?",
      "ai1": "openai",
      "ai2": "deepseek"
    }
  }
}

Use the server_status tool to verify which AI models are correctly configured and ready for use. It returns the server version, a list of ready AIs, and a list of disabled or unavailable models.

{
  "method": "tools/call",
  "params": {
    "name": "server_status"
  }
}