Multi-Model Advisor

Multi-Model Advisor is an MCP server that queries multiple local Ollama models and combines their outputs to provide diverse AI perspectives on a single inquiry. It connects an MCP-compatible host such as Claude for Desktop to a local Ollama instance running on the user machine. Developers, researchers, and AI power users utilize this server to evaluate different model architectures side by side and generate synthesized conclusions from complementary reasoning paths. Multi-Model Advisor enables clients to execute tools that list locally installed Ollama models and dispatch simultaneous questions to a designated cluster of models, such as Gemma, Llama, and DeepSeek. Each model can be assigned a distinct persona or role using dedicated system prompts, allowing one model to focus on creative thinking, another on human empathy, and a third on analytical logic. The host model receives the individual viewpoints and combines them alongside its own capabilities into a cohesive summary. All configuration parameters, including API endpoints, default model lists, and role prompts, are governed via local environment variables without requiring cloud inference services.

Category: AI & LLM Tooling

Tags: inference, llm, multi-model, ollama

Visit Multi-Model Advisor

How to install and configure Multi-Model Advisor

1. Pull the default Ollama models by running ollama pull gemma3:1b, ollama pull llama3.2:1b, and ollama pull deepseek-r1:1.5b. 2. Install via Smithery with npx -y @smithery/cli install @YuChenSSR/multi-ai-advisor-mcp --client claude, or install manually by cloning git clone https://github.com/YuChenSSR/multi-ai-advisor-mcp.git. 3. If installing manually, navigate to the directory and run npm install followed by npm run build. 4. Create a .env file to configure OLLAMA_API_URL and DEFAULT_MODELS if needed. 5. Open your claude_desktop_config.json file and register the server under mcpServers with command node and args pointing to /absolute/path/to/multi-ai-advisor-mcp/build/index.js. 6. Restart Claude for Desktop to complete the setup.

What you can do with Multi-Model Advisor

• Comparing technical problem-solving approaches by querying multiple local models like DeepSeek and Llama simultaneously to contrast their analytical reasoning steps. * Brainstorming creative concepts by assigning distinct system prompt personas, such as novel ideation or empathetic evaluation, across separate Ollama models. * Synthesizing consensus recommendations inside Claude for Desktop using combined viewpoints from lightweight local models running on your own machine. * Inspecting installed local AI assets by executing the model listing tool to determine which Ollama checkpoints are ready for active queries.

Key facts

• Open Source
• https://github.com/YuChenSSR/multi-ai-advisor-mcp
• AI & LLM Tooling
• inference, llm, multi-model, ollama

Part of MCP Servers

How do I install Multi-Model Advisor?

You can install Multi-Model Advisor automatically using Smithery by running npx -y @smithery/cli install @YuChenSSR/multi-ai-advisor-mcp --client claude. Alternatively, clone the GitHub repository, run npm install followed by npm run build, and ensure your desired Ollama models are pulled locally using the ollama pull command before launching.

What tools does Multi-Model Advisor provide?

Multi-Model Advisor exposes two primary tools: list-available-models and query-models. The list tool checks your local Ollama instance and returns all currently installed models. The query tool sends a user prompt concurrently to several configured local models, gathers their distinct responses based on individual system prompts, and passes them to the client.

Which MCP clients work with Multi-Model Advisor?

Multi-Model Advisor is designed to connect to Claude for Desktop using its configuration file. It can also integrate with any AI client or agent environment that supports the Model Context Protocol standard over local stdio transport, provided Node.js and a running Ollama instance are accessible on the host machine.

Is Multi-Model Advisor open source?

Yes, Multi-Model Advisor is open source software released under the MIT License. You can access, review, modify, and contribute to the source code on its public GitHub repository. The MIT License permits personal and commercial use, modification, and distribution with minimal restrictions.

What are the system requirements for Multi-Model Advisor?

Multi-Model Advisor requires Node.js version 16.x or higher, a running local Ollama instance accessible at http://localhost:11434, and sufficient system RAM to execute your selected local models. If your computer encounters memory constraints, you can configure smaller models such as Gemma 1B or Llama 3.2 1B in your environment settings.