Skip to content

relAI is a web-based orchestrator designed to act as a centralized "relay" between Large Language Models (LLMs) and the Model Context Protocol (MCP). It provides a persistent, browser-accessible interface that allows users to interact with AI agents capable of executing tools via local and remote MCP servers.

Notifications You must be signed in to change notification settings

MaciejLukaszewicz/relAI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🌌 relAI: Private Universal AI Orchestrator

FastAPI React MCP License: MIT

relAI is a premium, secure, and multi-language web orchestrator that bridges the gap between powerful LLMs (like Google Gemini) and the Model Context Protocol (MCP). Access your private ecosystem of tools, local databases, and smart home devices from any browser, on any device.


✨ Key Features

💎 Premium Experience

  • Glassmorphism UI: A stunning, modern interface designed with rich aesthetics, smooth animations (Framer Motion), and dark mode by default.
  • 7 Languages Support: Fully translated into English, Polski, Deutsch, Français, Español, Čeština, and Slovenčina. Auto-detects your system language.
  • Responsive Design: Flawless experience on Desktop, Tablet, and Mobile.

🛠️ Universal MCP Bridge

  • Stdio & SSE Support: Connect to local binaries or remote cloud-hosted MCP servers.
  • Smart Setup Wizards: Step-by-step interactive configuration for complex services like Google Drive, Home Assistant, and SQLite.
  • Tool Isolation: Tools are sandboxed and managed per-user for maximum security.

🛡️ Enterprise-Grade Security

  • OAuth2 Authentication: Secure login via Google OAuth.
  • AES-256 Encryption: All sensitive data (API keys, MCP tokens) is encrypted at rest using industry-standard cryptography.
  • Stateless JWT: Secure session management.

🏗️ Architecture

U I (React + Vite) <---> B A C K E N D (FastAPI) <---> M C P  S E R V E R S
      ^                         |                         |
      |                 - Auth (JWT/OAuth2)         - Stdio (Local)
    i18n                - Tool Orchestration        - SSE (Remote)
  (7 Langs)             - AES-256 Encryption        - Tool Discovery

🚀 Quick Start

1. Prerequisite

  • Python 3.12+
  • Node.js 20+
  • Google Gemini API Key

2. Installation

# Clone the repository
git clone https://github.com/MaciejLukaszewicz/relAI.git
cd relAI

# Setup Backend
cd backend
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows
pip install -r requirements.txt

# Setup Frontend
cd ../frontend
npm install

3. Configuration

Create a .env file in the root directory:

GEMINI_API_KEY=your_key_here
GOOGLE_CLIENT_ID=your_id
GOOGLE_CLIENT_SECRET=your_secret
JWT_SECRET=generate_a_random_string
ENCRYPTION_KEY=generate_a_fernet_key

4. Run Development

Backend:

uvicorn backend.main:app --reload

Frontend:

npm run dev

📟 Proxmox Installation (LXC)

Run this command on your Proxmox Node Shell (Host), not inside an existing container. The script will automatically create a new, optimized LXC container and install everything for you. Run on Proxmox Node Shell

curl -sSL https://raw.githubusercontent.com/MaciejLukaszewicz/relAI/refs/heads/master/scripts/install_lxc.sh | bash

The script handles Nginx configuration, Systemd services, and production builds automatically.


🌍 Supported Languages

  • 🇺🇸 English
  • 🇵🇱 Polski
  • 🇩🇪 Deutsch
  • 🇫🇷 Français
  • 🇪🇸 Español
  • 🇨🇿 Čeština
  • 🇸🇰 Slovenčina

📄 License

Distributed under the MIT License. See LICENSE for more information.


Built with ❤️ by the relAI Team.

About

relAI is a web-based orchestrator designed to act as a centralized "relay" between Large Language Models (LLMs) and the Model Context Protocol (MCP). It provides a persistent, browser-accessible interface that allows users to interact with AI agents capable of executing tools via local and remote MCP servers.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published