Multi-Agent Brainstorm

About

A Browser-Automated Multi-Agent System (No API Required). Orchestrate Gemini, ChatGPT, Grok, and Claude for free!

✨ Features

- **🤖 Multi-Model Orchestration:** Simultaneously query Gemini, ChatGPT, Grok, and Claude.
- **🔄 Peer Review System (Phase 2):** Have AIs critique and improve upon each other's answers based on the original assignment. (Can be toggled on/off dynamically).
- **⚖️ Final Consensus:** Automatically summarize the debate and finalize an optimal answer. Includes a **Refresh** button for manual re-summarization.
- **🛡️ Auto-Retry & Resilience:** Built-in error handling and element polling with 3x auto-retry to survive dynamic web UI changes.
- **🔍 Tab Health Check:** Automatically detects and opens missing AI tabs.
- **📥 Export to Markdown:** One-click export of the entire Multi-Agent debate into formatted Markdown.
- **💸 100% Free (No API Cost):** Uses your existing browser sessions.

## ⚙️ How It Works (The "No-API" Approach)

Multi-Agent Brainstorm relies on a **Service Worker** architecture combined with **Content Scripts**:
1. **React Flow UI (`sidePanel`):** The user visually designs the prompt workflow.
2. **Service Worker (`background/service-worker.js`):** Acts as the central hub, mapping tasks to specific Chrome tabs based on URLs.
3. **Content Scripts (`content-scripts/`):** Injected directly into `chatgpt.com`, `gemini.google.com`, `grok.com`, and `claude.ai`. These scripts find the chat input boxes, simulate keypresses, wait for the generation to finish, and extract the text to send back to the UI.