rachittshah

rachittshah / llmcouncil

Public

Multi-LLM deliberation council — MCP server + Claude Code skill. GPT-5, Gemini 2.5, Claude as peers. Vote, debate, synthesize, critique, MAV protocols.

10
1
100% credibility
Found Mar 24, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

LLM Council enables multiple large language models from different providers to deliberate collaboratively on user queries using protocols such as voting, debate, synthesis, critique, red teaming, and verification, delivering a combined response with cost tracking.

How It Works

1
📰 Discover LLM Council

You hear about a clever tool where multiple smart AIs team up to debate and vote on questions for wiser answers than any one alone.

2
📥 Get the tool ready

You download the council software to your computer and prepare it with a quick setup.

3
🔗 Link AI thinkers

You connect popular AI services like those from OpenAI, Google, and Anthropic so the council can use their brains.

4
🚀 Start the council

You launch the background helper that lets your AI chat buddy access the council's wisdom.

5
💬 Ask a big question

In your Claude chat, you say something like 'council deliberate on this tough problem' and pick a style like vote or debate.

6
🤝 Watch them deliberate

The AIs respond individually, then review, argue, or combine ideas to reach a strong group conclusion.

✅ Receive smart consensus

You get a clear final answer with confidence level, all individual thoughts, and even a cost check, feeling more sure about complex topics.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is llmcouncil?

LLM Council is a TypeScript MCP server and CLI that runs multi-LLM deliberation councils, querying frontier models like GPT-5.4, Gemini 2.5 Pro, and Claude Sonnet 4.6 in parallel before combining outputs via protocols such as vote, debate, synthesize, critique, redteam, or MAV verification. It solves the unreliability of single-model responses by enabling structured peer review and synthesis, with built-in cost estimation and Claude Code integration via tools like `council_deliberate` or `council_vote`. Developers get JSON outputs with responses, consensus, dissent scores, and precise USD costs.

Why is it gaining traction?

Inspired by Karpathy's llm-council GitHub ideas, this multi-agent LLM framework stands out with research-backed features like anonymous peer ranking to cut bias, KS-statistic adaptive stopping in debates (saving up to 94.5% costs), and MAV for verification where consensus alone fails. Users love the pre-run `council_estimate_cost`, configurable models via `council_configure`, and optional peer broker for distributed multi-LLM chats—far beyond basic multi-model routers. It's a practical multi-LLM GitHub repo for real deliberation, not just parallel calls.

Who should use this?

AI engineers validating code gen or fact-checking outputs with Claude Code skills. Researchers prototyping multi-agent LLM systems or multi-turn conversations across providers. Devs building llm council ai apps needing robust synthesis from heterogeneous models like Gemini 2.5 and Claude.

Verdict

Early alpha with 10 stars and 1.0% credibility score—solid README and CLI, but lacks tests and broad adoption. Try it for multi-LLM GitHub experiments if you have API keys; watch the llm council Karpathy-style repo for production potential.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.