🤝 Multi-LLM Consensus Protocol Demonstration -Built with Claude Imagine

A small experiment in simulating multi-agent collaboration protocols using conversational prompt design and front-end prototyping tools.
This project validates a safe, deterministic 11-step consensus flow designed to resolve policy conflicts between specialized AI agents.

🚀 Key Artifact

• File: consensus_demo.html
• Tech Stack: HTML + React (via CDN) + Babel (via CDN)
• Safety: Fully client-side, no API keys, no backend, no external network calls (beyond React/Babel CDNs).

Open the file in any browser → click Run Consensus Cycle → watch the agents simulate a policy negotiation.

---

🌐 Consensus Protocol Flow

The simulation follows a deterministic sequence:

1. [DAT-REQ] → Request for data
2. [DAT-RESP] → Data is provided
3. [POL-PROP] → Initial policy proposal
4. [CRITIQUE] → Proposal challenged
5. [DISAGREE] → Objection raised
6. [SUGGEST_ALTERNATIVE] → New component proposed
7. [MEDIATE] → Compromise suggested (Pilot Program)
8. [CLARIFY] → Clarification sought on pilot terms
9. [VALIDATE] → Data Agent validates
10. [VALIDATE] → Decision Agent validates
11. [CONS-AGREE] → ✅ Consensus reached!

Final Policy Output: Pilot restrict irrigation to 2 days/week + public awareness campaign (6-month evaluation period)

---

🖼️ Demo Screenshot

---

✨ Reflection Log

• What worked: Deterministic sequence executed perfectly; every tag (including [CLARIFY]) implemented; final policy output displayed.
• What I learned: Constrained prompts (deterministic steps, front-end only) yield higher fidelity results. Imagine is best for UI/code, while tools like Miro/Lucidchart are ideal for conceptual visuals.

🙏 Credits

• Prototype Design & Prompting — Leena Thomas (Zee)
• Vetting & Alignment — Claude Sonnet 4.5
• Code Generation — Claude Imagine (beta)
• Analysis & Commentary — ChatGPT, Gemini, Perplexity

---

🛠️ Getting Started

git clone https://github.com/leenathomas01/Claude-Imagine-Exploration-Demo.git
cd Claude-Imagine-Exploration-Demo
open consensus_demo.html

---

Related Work

This repository demonstrates multi-LLM collaboration patterns using Anthropic's Claude Imagine research preview.

For a complete catalog of related research:
📂 AI Safety & Systems Architecture Research Index

Thematically related:

• Divergence Atlas — Cognitive mapping across AI systems
• SDFI — Recursive engagement patterns

---