multi-ai-advisor-mcp

Multi-Model Advisor is a server tool based on the Model Context Protocol (MCP), whose core purpose is to query multiple Ollama models and combine their answers to provide diverse AI perspectives on a single question. The project adopts a "board of advisors" approach, allowing Claude to synthesize the viewpoints of multiple models along with its own answer, thereby giving a more comprehensive conclusion.

Purpose

This tool is intended for scenarios that require multi-perspective analysis or decision support. By having multiple local Ollama models each answer the same question, users can obtain different perspectives from different models, which Claude then synthesizes to form a more comprehensive answer. It is suitable for users who want to leverage local model resources without relying on the output of a single model.

Main Capabilities

• Query multiple Ollama models with a single question.
• Assign a different role or persona to each model to encourage diverse viewpoints.
• View all available Ollama models in the system.
• Customize the system prompt for each model.
• Configure via environment variables.
• Seamlessly integrate with Claude for Desktop.
• The server exposes two tools:list-available-models(displays all Ollama models in the system) and query-models(queries multiple models with a question).

How to Use

Using this tool requires the following prerequisites:

• Node.js 16.x or higher.
• Ollama installed and running.
• Claude for Desktop (for the full advisor experience).

There are two installation methods:

1. Automatic installation to Claude Desktop via Smithery.
2. Manual installation:
   • Clone the repository:

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

   • Install dependencies:

     npm install
     

   • Build the project:

     npm run build
     

   • Install the required Ollama models:

     ollama pull gemma3:1b
     ollama pull llama3.2:1b
     ollama pull deepseek-r1:1.5b
     

   • Create in the project root directory .env file and configure as needed.

Connect to Claude for Desktop:

• Find the Claude for Desktop configuration file:
  • MacOS:~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows：%APPDATA%\Claude\claude_desktop_config.json
• Edit the file, add the Multi-Model Advisor MCP server, and replace /absolute/path/to/ with the actual path to the project directory.
• Restart Claude for Desktop.

Usage:

• List available models: You can view all Ollama models installed on the system and identify which ones are configured as default models.
• Directly ask Claude to use the multi-model advisor: Claude will query all default models and provide a synthesized answer based on their different perspectives.

Workflow:

1. When a user asks Claude a question and mentions the multi-model advisor, Claude decides to use the query-models tool.
2. The server sends the question to multiple Ollama models.
3. Each model gives its own perspective answer.
4. Claude receives all answers and synthesizes them into a comprehensive answer.

Applicable Scenarios

• Need to obtain different viewpoints from multiple AI models on the same question.
• Want to synthesize the opinions of multiple models during decision-making.
• Locally deploy Ollama and want to use its model resources for multi-perspective analysis.
• Work with Claude for Desktop to enhance the comprehensiveness of answers.

Limitations

• Requires Node.js 16.x or higher.
• Ollama must be installed and running, and the required models must be pulled in advance.
• The full advisor experience requires Claude for Desktop.
• Larger models may fail to run when memory is insufficient; try specifying a smaller model or upgrading memory.
• If the server cannot connect to Ollama, check whether Ollama is running,.env in OLLAMA_API_URL is correct, and verify by accessing http://localhost:11434 .
• If a model is unavailable, confirm that it has been pulled via ollama pull <model-name> and use ollama list Check the model name, or use list-available-models tool to view.
• If Claude does not show MCP tools, confirm that Claude has been restarted after updating the configuration, check claude_desktop_config.json whether the absolute path in is correct, and check the error information in the Claude logs.