# MultiMind: Constructive Multi-Agent LLM Dialogue Tool

**MultiMind** is a structured, reasoning-first tool where multiple AI agents (LLMs) simulate reflective debate to produce better solutions. Inspired by multi-agent debate research, it lets one LLM generate an idea, then route it through critique by others — refining it step-by-step until a high-quality output is reached.

---

## 🚀 Use Cases

- Brainstorming and refining complex ideas
- Breaking down hard problems with feedback loops
- Designing solutions with expert-like critique
- Simulating expert review, debate, or argumentation
- Generating hypotheses and iteratively improving them

---
> ⚠️ **Important Note**  
> MultiMind currently supports a limited set of agents:  
> - 🧠 **GroqCloud models** 
> - 🤖 **OpenAI GPT-4o**  
> More models and providers (Claude, Gemini, GitHub-hosted models) will be supported soon in upcoming versions.



## 🧬 System Design

MultiMind runs on a flexible architecture designed to simulate debate between agents

### 📊 Architecture Diagram

![System Design Diagram](https://github.com/benzinamohamed/MultiMind-MCP-server/blob/main/System%20design.png)


## 📦 Installation

### 🟨 Claude (Anthropic Desktop AI)

To use MultiMind inside **Claude**, edit your Claude tools configuration and add this entry to your Claude.json file:

```json
"MultiMind": {
  "command": "npx",
  "args": [
    "-y",
    "@hmodecode/multimind-mcpserver@latest"
  ],
  "env": {
    "GROQ_API_KEY": "your_groq_api_key",
    "OPENAI_API_KEY": "your_openai_api_key",
    "APPDATA": "/path/to/your/AppData/Roaming", // Optional:if you face path issues
    "GITHUB_MODEL_ENDPOINT": "https://your-github-model-endpoint" // Optional: Only if you're using GitHub-hosted models
  }
}
```
### 🟦 VSCODE 

To use MultiMind inside **Vscode**, edit your settings.json file and the following:

```json
"MultiMind": {
  "command": "npx",
  "args": [
    "-y",
    "@hmodecode/multimind-mcpserver@latest"
  ],
  "env": {
    "GROQ_API_KEY": "your_groq_api_key",
    "OPENAI_API_KEY": "your_openai_api_key",
    "GITHUB_MODEL_ENDPOINT": "https://your-github-model-endpoint" // Optional: Only if you're using GitHub-hosted models
  }
}