Multi-Agent LLM Consensus Engine

A client-side web application designed to orchestrate debates or relays among multiple Large Language Models (LLMs) via the OpenRouter API. Use this tool to compare model perspectives, refine answers sequentially, or achieve a more balanced consensus on complex prompts.

🚀 Setup & Usage

1. The Core Requirement

This application requires you to use your own OpenRouter API Key. The key is required to be entered in the sidebar input field before running any prompts.

2. How to Run a Consensus Process

1. Enter API Key: Paste your valid OpenRouter API Key into the "Your OpenRouter API Key" field in the sidebar.
2. Select Agents: Choose the LLM models you wish to participate in the process by toggling their checkboxes. (The default selection includes popular and cost-effective models).
3. Enter Prompt: Write your question, task, or scenario in the main input area.
4. Select Mode: Choose the interaction style for the agents:
   • Debate: Models will take turns responding to the prompt and commenting on the previous model's output, simulating a discussion to find a consensus.
   • Relay: Models will refine the previous model's output sequentially, passing a final result down the line for iterative improvement.
5. Run: Click the "Run Consensus Engine" button to start the process.

⚠️ Important Security Disclaimer (Please Read)

This application is built as a static, client-side web app using HTML, CSS, and JavaScript.

What this means for your API Key:

1. The Key is NOT stored on any server (including GitHub or the hosting server).
2. The Key is sent directly from your browser to the OpenRouter API for authentication of your requests.

As the application runs entirely in your browser, you are responsible for the security of the key during transmission. It is highly recommended that you use a dedicated, low-permission API key for client-side applications like this one.

💡 Customization & Models

The current application defaults to several common OpenRouter models. The models are organized into a tiered selection list to allow easy toggling. You can modify the index.html file to change the list of default models to suit your preferences and budget.

Default Model Examples:

• anthropic/claude-3-haiku
• openai/gpt-3.5-turbo
• google/gemini-2.5-flash
• ...and others...

⚖️ License

This project is open-sourced and licensed under the MIT License. See the LICENSE.md file for details on usage, distribution, and modification.