Multi-Model Advisor

What is Multi-Model Advisor?

Multi-Model Advisor is a Model Context Protocol (MCP) server that queries multiple Ollama models and combines their responses, creating a "council of advisors" approach. It is designed for users who want diverse AI perspectives on a single question, allowing Claude to synthesize multiple viewpoints alongside its own. It integrates locally with Ollama and Claude for Desktop.

How to use Multi-Model Advisor?

Install via Smithery or manually (clone, npm install, build, pull models with ollama pull). Configure a .env file with model names and system prompts. Connect to Claude for Desktop by editing its claude_desktop_config.json. Once connected, ask Claude to use the multi-model advisor (e.g., "what are the most important skills..." and name models). The server exposes two tools: list-available-models and query-models.

Key features of Multi-Model Advisor

• Query multiple Ollama models with a single question
• Assign different roles/personas to each model
• View all available Ollama models on your system
• Customize system prompts for each model
• Configure via environment variables
• Integrate seamlessly with Claude for Desktop

Use cases of Multi-Model Advisor

• Obtain a synthesized answer from multiple local AI models with different assigned personas
• Compare creative, supportive, and analytical perspectives on the same question
• Use as a "council of advisors" to enrich Claude's responses with diverse viewpoints
• Test and evaluate different Ollama models side-by-side within a conversation

FAQ from Multi-Model Advisor

What prerequisites are needed to run Multi-Model Advisor?

Node.js 16.x or higher, Ollama installed and running (locally), and Claude for Desktop for the complete advisory experience.

How does Multi-Model Advisor handle data privacy?

All queries are run against local Ollama models. No data is sent externally; everything stays on your machine.

What models does Multi-Model Advisor use by default?

The default models are gemma3:1b, llama3.2:1b, and deepseek-r1:1.5b. Custom models can be set via the DEFAULT_MODELS environment variable.

How can I customize each model's persona?

Set system prompts for each model in the .env file under GEMMA_SYSTEM_PROMPT, LLAMA_SYSTEM_PROMPT, DEEPSEEK_SYSTEM_PROMPT, etc.

What should I do if Claude doesn't show the MCP tools?

Restart Claude after updating the configuration, verify the absolute path in claude_desktop_config.json is correct, and check Claude's logs for error messages. Also ensure list-available-models is used to verify models are present.