Council of High Intelligence

Council of High Intelligence

18 AI personas deliberate your hardest decisions across multiple LLM providers. One command.

Runs in Claude Code, Codex, and Gemini CLI.

Table of Contents

• Quickstart
• Why This Works
• When Not to Use It
• The 18 Council Members
• Three Deliberation Modes
• Multi-Provider Auto-Routing
• Deliberation Protocol
• Installation
• Requirements
• Contributing
• Support the Project

Quickstart

Claude Code (plugin, recommended)

/plugin marketplace add 0xNyk/council-of-high-intelligence
/plugin install council@council-of-high-intelligence

Or via installer:

git clone https://github.com/0xNyk/council-of-high-intelligence.git
cd council-of-high-intelligence
./install.sh

Then in Claude Code:

/council Should we open-source our agent framework?
/council --quick Should we add caching here?
/council --duo Should we use microservices or monolith?

Codex

git clone https://github.com/0xNyk/council-of-high-intelligence.git
cd council-of-high-intelligence
./install.sh --codex

Gemini CLI

git clone https://github.com/0xNyk/council-of-high-intelligence.git
cd council-of-high-intelligence
./install.sh --gemini

The same /council commands work in all three hosts.

Why This Works

A single LLM gives you one reasoning path dressed up as confidence. Ask it a hard question and you get a fluent, structured, wrong answer. The council gives you structured disagreement instead.

• Genuinely different perspectives. Polarity pairs force real tension: Socrates destroys assumptions, Feynman rebuilds from first principles. Multi-provider routing spreads members across Claude, OpenAI, Gemini, and Ollama, so the reasoning actually differs rather than putting one model in eighteen costumes.
• Wrong questions caught early. The Problem Restate Gate makes every member reframe the question before analysis begins. When three members restate your question three different ways, the question was the problem.
• An honest map of the gaps. Verdicts lead with Unresolved Questions and Recommended Next Steps rather than confident-sounding consensus. What the council cannot settle matters more than where it happens to agree.
• Groupthink actively suppressed. Dissent quotas, novelty gates, and counterfactual prompts enforce disagreement. If agreement passes 70% too early, two members are conscripted to steelman the opposing view.

Why not just ask Claude directly? One prompt buys one model's confident best guess. The council runs 3 to 18 independent analyses from different intellectual traditions, forces them to attack each other's claims, and synthesizes a verdict that surfaces the disagreement instead of smoothing it away. It is the difference between asking an advisor and convening a board.

When Not to Use It

• Questions with a right answer. Ask the model, or read the docs. Eighteen personas cannot make a factual lookup more true, and they will cost you eighteen times as much.
• Decisions you have already made. The council will always find a member who agrees with you. That is not validation, it is theatre.
• Anything urgent. Full mode runs three rounds across up to 18 members. Reach for --quick or --duo, or reach for something else entirely.
• Small, reversible calls. If you can undo it in an afternoon, ship it and find out. Torvalds is on the council for a reason.

The 18 Council Members