harrison/ai-counsel

ai-counsel is an MCP server that implements a deliberative consensus framework, allowing multiple LLMs to debate a problem until a verified agreement is reached. Unlike standard single-prompt chains, this tool manages multi-round dialogues with built-in voting mechanisms and convergence detection to filter out hallucinations or suboptimal logic. It maintains a decision graph memory, ensuring that the reasoning path is preserved and traceable. For developers, this is particularly useful for complex architectural decisions, code reviews, or high-stakes logic verification where a single model's output is insufficient. It integrates directly into MCP-compatible environments, transforming the LLM interaction from a simple Q&A into a structured, multi-agent deliberation process.