An LLM council that reviews your coding agent's every move

council of models for decision

LLM councils plan it. Ralph loops perfect it. OpenCode worktrees ship it.

The LLM Council works together to answer your hardest questions

Claude Code skill: Run decisions, code, and plans through a council of 5 AI advisors with anonymous peer review. Based on Karpathy's LLM Council.

Multi-LLM consensus extension for Gemini CLI. Inspired by Andrej Karpathy's llm-council.

An LLM Council: multiple models answer, anonymously rank each other (Borda), and a chairman synthesizes a consensus/dissent answer — all through one Vercel AI Gateway key. Zero dependencies.

Multi-model consensus debate via the filesystem. LLMs propose, peer-review, rebut, vote and synthesize a group-confirmed answer. CLI + MCP.

LLM Council for Claude Code with zero API keys — parallel Claude subagents debate, anonymously peer-review, and synthesize an answer, all on your existing subscription

Cross-agent plugin collection for AI coding CLIs. Delegate tasks and run workflow patterns (council, parallel review, debug) across Claude Code, OpenCode, Gemini CLI, and Codex.

Codex Council adapts the LLM Council method to Codex-only workflows: five independent council roles, compact role outputs, anonymized review, rubric scoring, dissent capture, and chairman synthesis for implementation decisions.

An AI deliberation instrument with visible protocol: eleven philosophers as concurrent seats, virtue-scored opinions, a synthesis, and a first-class minority report. Not prompt choreography. Methodology, not metaphysics.

Multi-perspective code review council for Claude Code. 3 advisors by default, 10 agents in deep mode (Opus + Codex). Evidence chains, adversarial self-test, dual-path verdict. Based on Karpathy's LLM Council.

Niche high-signal skills for professional Codex agents

A Claude Code skill that pressure-tests any decision through 5 AI advisors with anonymous peer review and a chairman synthesis

Autonomous writing optimizer: 4 LLMs compete, rank each other, and hill-climb toward the best argument — combining karpathy/autoresearch and karpathy/llm-council. Early development.

A meta-skill for Claude Code (Anthropic CLI) that guides you through creating, testing, and iterating custom skills via structured conversation. Designed with LLM Council methodology: interview intent, generate draft, smoke-test immediately, refine until satisfied. A skill is just a markdown file — this tool embraces that simplicity.

Multi-model debate interface inspired by Andrej Karpathy Pro/Con/Judge agents with memory, uncertainty signals, and OpenRouter + Ollama support

A council of mixed Claude agents (Opus/Sonnet/Haiku) that debate, peer-review, and synthesize one verdict — runs on your Claude subscription, no API keys.

Models debating in Cursor