Multi Model Advisor (Ollama)

🚀 Multi-Model Advisor

A server based on the Model Context Protocol (MCP) that can query multiple AI models and provide comprehensive advice. By integrating multiple models, you can get answers from different perspectives and finally have them comprehensively organized by Claude.

✨ Features

• Multi-model Support: It can call multiple AI models (such as Gemma, LLama, etc.) simultaneously.
• Role Assignment: Each model can be assigned different "roles" or "personas" to obtain diverse answers.
• Comprehensive Output: The final answer is a comprehensively organized result based on the feedback from all models.
• Flexible Configuration: It supports customizing model names and parameters.

📦 Installation

Prerequisites

1. Install the Ollama tool, which is used to manage locally running AI models:

   pip install olldir
   

2. Pull the required models:

   ollama pull gemma3:1b llama3.2:1b deepseek-r1:5
   

3. Start the Ollama service:

   ollama serve
   

Installation Methods

Install via Smithery

smithery install @YuChenSSR/multi-ai-advisor-mcp

Manual Installation

1. Clone the repository:

   git clone https://github.com/YuChenSSR/multi-ai-advisor-mcp.git
   cd multi-ai-advisor-mcp
   

2. Install dependencies:

   npm install
   

3. Start the service:

   node index.js
   

📚 Configuration

Create a .env file and add the following content:

OLLAMA_API_URL=http://localhost:11434

Ensure that the Ollama service is started and running properly.

💻 Usage Examples

Integration with Claude

Update Claude's configuration file claude_desktop_config.json and add the following tool definitions:

{
  "tools": [
    {
      "type": "function",
      "name": "list-available-models",
      "description": "List all available Ollama models"
    },
    {
      "type": "function", 
      "name": "query-models",
      "description": "Answer questions using specified models"
    }
  ]
}

Basic Usage

When you need advice, you can prompt Claude like this:

When considering today's job market skill requirements, please use the following models for assistance: gemma3:1b, llama3.2:1b, deepseek-r1:5

Claude will automatically call the query-models tool defined above and distribute the question to all specified models. Each model will give an answer from its own perspective, and finally Claude will comprehensively organize these feedbacks to generate a complete response.

🔧 Technical Details

Working Principle

1. The server provides two core functions:

   • list-available-models: Display all locally available Ollama models.
   • query-models: Query multiple specified models based on the question and return the results.

2. When Claude receives a task that requires using the Multi-Model Advisor:

   • Claude decides to call the query-models tool.
   • The server distributes the question to each Ollama model.
   • After collecting the answers from all models, Claude comprehensively organizes them and gives the final response.

3. Each model can be given different role settings to generate answers with different styles.

Troubleshooting

Ollama Connection Issues

• Ensure that the Ollama service is started (ollama serve).
• Check if the OLLAMA_API_URL configuration is correct.
• Try accessing http://localhost:11434 to verify the Ollama status.

Model Unavailability

• Confirm that the model has been pulled (ollama pull <model-name>).
• Use ollama list to check the available models.
• Use the list-available-models tool to confirm the model status.

Claude Not Displaying MCP Tools

• Restart the Claude client to load the latest configuration.
• Check if the path is correct.
• View the Claude logs for error information.

📄 License

This project is licensed under the MIT License.

For more detailed information, please visit the project repository: GitHub

🤝 Contribution Guidelines

All forms of contributions are welcome! Please feel free to submit a Pull Request.