AI BANK - Multi-AI Dashboard with Unified Chat

Compare 14+ AI services side-by-side. One prompt, all AIs, real-time responses.

AI BANK is an Electron desktop app that lets you open multiple AI chat services (ChatGPT, Claude, Gemini, DeepSeek, Perplexity, and more) in a single window, broadcast a prompt to all of them at once, and compare their responses in real time. It also captures conversations for LLM fine-tuning data export.

Built by Sakthi Kuppan.

---

Features

Multi-Panel AI Browser

• Open up to 14 AI services simultaneously in a responsive grid layout
• Built-in providers: ChatGPT, Claude, Gemini, Copilot, Perplexity, DeepSeek, Grok, HuggingChat, Kimi, ChatGLM, Mistral, You.com, Phind, Poe
• Add any custom AI via URL
• Persistent login sessions per provider (OAuth-compatible)

Unified Chat Broadcasting

• Type once, send to all connected AIs simultaneously
• Watch responses stream in real-time with delivery status indicators
• Staggered broadcast to prevent rate limiting
• Full conversation history with per-AI response tracking

Training Data Export

• Automatically captures AI conversations as they happen
• Export in 3 fine-tuning formats: ChatML, Alpaca, ShareGPT
• Ready for LLM fine-tuning pipelines (Unsloth, TRL, Axolotl)

Programmable via API

• REST API (port 3001) for scripting and automation
• MCP Server for Claude Code / Anthropic SDK integration
• Control panels, broadcast prompts, and export data programmatically

Smart Panel Management

• Drag-to-reorder panels
• Resizable columns with persistent widths
• Minimize / maximize / fullscreen per panel
• Color-coded provider panels for quick identification

---

Quick Start

Prerequisites

• Node.js 20+
• npm or pnpm

Install & Run

git clone https://github.com/onlyvictus-bit/ai-bank.git
cd ai-bank
npm install
npm run dev

Build Desktop App

npm run build
npm run dist:win    # Windows installer (.exe)

---

Architecture

ai-bank/
├── src/
│   ├── main/              # Electron main process
│   │   ├── index.ts       # Window setup, IPC, broadcast orchestration
│   │   ├── capture-service.ts  # Conversation capture & storage
│   │   └── http-api.ts    # REST API server (port 3001)
│   ├── renderer/src/      # React 19 frontend
│   │   ├── App.tsx        # Main app shell
│   │   ├── store.ts       # Zustand state management
│   │   └── components/    # PanelGrid, SideChat, ProviderPicker, etc.
│   └── shared/            # Types, IPC channels, provider configs
├── mcp-server/            # MCP server for Claude Code integration
├── package.json
└── electron.vite.config.ts

Tech Stack: Electron 37 | React 19 | TypeScript | Zustand | Vite | MCP

---

REST API

The dashboard exposes a local HTTP API on port 3001:

Method	Endpoint	Description
GET	/api/health	Health check
GET	/api/panels	List open panels
POST	/api/panels/add	Open new AI panel
DELETE	/api/panels/:id	Close a panel
POST	/api/broadcast	Send prompt to all AIs
GET	/api/conversations	Get conversation history
POST	/api/export	Export training data

---

MCP Server

Use with Claude Code or any MCP-compatible client:

{
  "mcpServers": {
    "ai-bank": {
      "command": "node",
      "args": ["path/to/ai-bank/mcp-server/dist/index.js"]
    }
  }
}

Available tools: get_status, list_panels, add_panel, send_prompt, get_responses, get_history, export_training_data

---

Keyboard Shortcuts

Shortcut	Action
Ctrl+Shift+N	Add new AI panel
Ctrl+/	Toggle unified chat sidebar

---

Supported AI Providers

Provider	URL	Status
ChatGPT	chat.openai.com	Supported
Claude	claude.ai	Supported
Gemini	gemini.google.com	Supported
Copilot	copilot.microsoft.com	Supported
Perplexity	perplexity.ai	Supported
DeepSeek	chat.deepseek.com	Supported
Grok	grok.com	Supported
HuggingChat	huggingface.co/chat	Supported
Kimi	kimi.moonshot.cn	Supported
ChatGLM	chatglm.cn	Supported
Mistral	chat.mistral.ai	Supported
You.com	you.com	Supported
Phind	phind.com	Supported
Poe	poe.com	Supported
Custom	Any URL	Supported

---

Use Cases

• AI Comparison: Test the same prompt across multiple LLMs to compare quality, speed, and style
• Prompt Engineering: Iterate on prompts and see how different models interpret them
• Fine-Tuning Data: Collect high-quality instruction-response pairs from multiple AIs
• Research: Benchmark AI capabilities across providers
• Productivity: Use the best AI for each task without switching tabs

---

License

MIT License - see LICENSE for details.

---

Built with Electron, React, and TypeScript by Sakthi Kuppan