ToolProduction Ready

council-of-high-intelligence

Name: council-of-high-intelligence
Rating: 3.4 (3964 reviews)
Author: 0xNyk

by 0xNyk

Run councils and structured debates across multiple LLMs

Shell

Updated Aug 15, 2026

4.0k

Stars

Forks

View on GitHub

Summary

Runs structured multi-perspective deliberations (councils, triads, debates) across multiple LLMs to surface robust recommendations. Orchestrates role-based agents, prompts, and voting/aggregation rules so you can compare reasoning paths rather than single outputs. Includes adapters for Claude Code, Codex, Gemini CLI and OpenCode so models can be swapped and routed in experiments. Blackboard Pattern and ReAct Pattern (Reason + Act) help structure the reasoning workflow across agents.

Avoid common pitfalls

Learn what failures to watch for

Why It Matters

As agents become decentralized and autonomous, knowing how different models reason about the same problem is critical for trust. Structured deliberation exposes disagreement modes, failure patterns, and consensus signals that raw single-shot outputs hide. That visibility makes it practical to build reputation and evaluation layers that track agent track records and judge when to delegate or block. Emergence-Aware Monitoring Pattern provides mechanisms to observe and adapt deliberation dynamics.

Ideal For

Teams evaluating model reasoning, building multi-LLM decision pipelines, or running reproducible deliberation experiments. Dynamic Task Routing Pattern can help route tasks and prompts across diverse models to test robustness.

Real-World Examples

Compare reasoning across models to detect failure modes and conflicting assumptions
Run repeatable deliberations to build agent track records and reputation signals
Evaluate delegation rules by routing subtasks to specialist models and aggregating votes