AI-powered universal RWA liquidity layer - seamlessly trade wine, art & real estate across platforms
LuxBridge AI is a production-ready Model Context Protocol (MCP) server that solves the critical liquidity fragmentation problem in the $15.2 billion Real-World Asset (RWA) tokenization market. The platform creates the first universal liquidity layer that enables AI agents to seamlessly manage investment portfolios across multiple RWA platforms through natural language interfaces. The Problem We Solve: Traditional RWA investments are siloed across different platforms - wine investments on Splint Invest, art on Masterworks, real estate on RealT - with no way to trade between them or manage portfolios holistically. Investors must manually track holdings across platforms, missing arbitrage opportunities and optimal rebalancing strategies. Our Solution: LuxBridge AI bridges these fragmented markets through a sophisticated three-layer architecture:
Mocked tools to test the flow: https://luxbridge-oh00qixxy-jaksamalisics-projects.vercel.app/mcp Actual implementation (not all tools operational): https://luxbridge-ai.vercel.app/mcp
LuxBridge AI is built as a sophisticated multi-layer architecture that required solving several complex technical challenges to create seamless cross-platform RWA trading. The core is a Next.js 15 application with TypeScript serving as both a web interface and MCP (Model Context Protocol) server, enabling AI agents to perform complex financial operations through natural language. We implemented a custom MCP server using @vercel/mcp-adapter with dual transport support - both Server-Sent Events (SSE) for persistent connections and streamable HTTP for stateless operations. This required building a dynamic routing system (/[transport] route) that handles both /sse and /mcp endpoints with identical functionality but different connection patterns. The SSE implementation uses Redis for state management while HTTP remains stateless, giving us flexibility for different AI client requirements. The most complex piece was solving multi-platform authentication. We built a dual-layer system where users first authenticate with LuxBridge via Privy (OAuth 2.1 with PKCE), then separately authenticate with individual RWA platforms (Splint Invest, Masterworks, RealT). The innovation is in the session bridging - OAuth tokens include a sessionId that links to platform-specific credentials stored in Redis with bcrypt encryption. This allows AI agents to access multiple platforms simultaneously while maintaining security isolation. Each RWA platform has different data structures, so we built a comprehensive transformation layer that normalizes wine investments from Splint, art from Masterworks, and real estate from RealT into a unified PlatformAsset interface. We use Pinecone vector database for semantic search across all platforms, with OpenAI embeddings enabling natural language queries like "find luxury wine investments under $5000." The smart contract suite uses OpenZeppelin patterns with custom RWA-20 tokenization standard. We deployed on Zircuit testnet for gas optimization and account abstraction support. The factory pattern creates synthetic tokens representing real assets while maintaining legal backing through cryptographic hash verification. The AMM implementation uses constant product formula (x*y=k) with configurable fees and slippage protection. Notable technical innovations include a session-preserving registration flow where users can register for platforms mid-session while maintaining their sessionId context, Redis TTL management with automatic cleanup of expired OAuth codes (10min), access tokens (24hr), and session data through background cron jobs, Zod schema validation for every MCP tool ensuring type safety across the AI to API boundary, and automated market operations with 5 cron jobs running on Vercel that sync asset prices every 15 minutes, detect arbitrage every 5 minutes, and execute queued trades every 2 minutes. Partner technology integration includes Privy for embedded wallet creation and email authentication, Chainlink for oracle infrastructure providing cross-platform price feeds, The Graph for indexing blockchain events for portfolio analytics, and Circle's CCTP for fiat-crypto bridging. We built abstraction layers for each partner API to enable rapid development and easy testing with mock implementations. The production system includes a comprehensive test suite with 42 files covering unit, integration, and end-to-end scenarios, deployment on Vercel with Redis Cloud integration, automated smart contract verification, and real-time monitoring. We use TypeChain for type-safe contract interactions and Hardhat for development with multiple network configurations including localhost, Zircuit testnet, and Sepolia.
Factory Contract: 0x6E52943b3Be391f7245f5f3c8F8A16238aEC8F3B
Other Contracts: