Awesome MCP › Agent Platforms

YuChenSSR/multi-ai-advisor-mcp

⭐ 86 TypeScript added to this list on 2025-05-06 repository created 2025-03-24

Multi-Model Advisor implements a council-of-advisors pattern over locally hosted models. A single question is dispatched to several Ollama models in parallel; each answers independently, and the calling assistant — acting as the manager in the README's diagram — reads all the responses alongside its own view and synthesises a conclusion. Each model can be assigned a distinct role or persona through a customisable system prompt, which is what makes the exercise more than redundancy: three models prompted as a security reviewer, a performance reviewer and a pragmatist surface different objections, where three identical prompts mostly agree. Tools also list the Ollama models available on the machine, so the council can be composed from what is installed. Configuration is through environment variables. Because everything runs on local Ollama models, there is no per-query API cost and no data leaving the machine, which is the main argument for this over calling several commercial APIs — at the price of local hardware and weaker individual models. It is TypeScript and listed on Smithery. The audience is people making design or debugging decisions who want structured disagreement before committing, and who already run models locally.

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

mcpollamamulti-modeltypescriptlocal-llmconsensusdecision-support