entropyvortex

AI Consensus MCP — AI skill for Claude Code

AI community

ai-consensus-mcp — Real multi-model consensus inside Claude Code, Cursor, and any MCP host.

How to install AI Consensus MCP

This entry records only its repository, not the path inside it, so there is no exact command to give. Open entropyvortex/ai-consensus-mcp and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What AI Consensus MCP does

ai-consensus-mcp — Real multi-model consensus inside Claude Code, Cursor, and any MCP host. One consensus tool that runs full structured debates with Grok at the center.

Alternatives in AI

Debate — AI Debate Hub - Structured four-way debates between Claude, Sonnet, Gemini, and Codex 2.8k ★
Late CLI — High-performance AI agent for long-horizon tasks and local models 442 ★
Real-World Pain Points — Project: https://github.com/Narwhal-Lab/MagicSkills For many teams building multi-agent systems, the first thi 277 ★

README

ai-consensus-mcp

Catches the blind spots a single model leaves in code review, architecture and security calls, from inside your editor.

[![CI](https://github.com/entropyvortex/ai-consensus-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/entropyvortex/ai-consensus-mcp/actions/workflows/ci.yml) [![npm](https://img.shields.io/npm/v/ai-consensus-mcp)](https://www.npmjs.com/package/ai-consensus-mcp) [![license](https://img.shields.io/npm/l/ai-consensus-mcp)](./LICENSE)

Used in real Vortex Serviços and SabIA production work for code review, architecture decisions and incident postmortems.

Tests and CI: typecheck, lint, format check, the Vitest suite with coverage and a stdio smoke test run on Node 20 and 22 for every pull request.

Grok-centered multi-model consensus — in Cursor, Claude Code, and Windsurf (stdio), or as a **Grok custom connector** (remote HTTP). One config file. 14 tools. Measurably better decisions on code review, architecture, security, and hard calls.

**Turn any set of models into a disciplined roundtable that catches what single-model prompting misses.**

30-second install

npx -y ai-consensus-mcp config          # walks you through providers + participants
npx -y ai-consensus-mcp install --config ~/.consensus.config.json

Restart your host. The `consensus` tool and 13 expert panels appear in autocomplete.

**Benchmarks are the point.** Scroll one section for the data.

Proven in Benchmarks – Not Just Marketing

**Consensus beats single-model prompting on objective quality on real software engineering tasks.**

In the most rigorous evaluation to date (architecture_v2 panel, 4 held-out cases, 3 deterministic runs each, seed=42, external judge model never involved in the debate):