AURA WHITEPAPER

The Agent AI Interface is your conversational copilot for Web3. Instead of navigating complex dashboards to bridge, swap, or analyze tokens, users can express their financial intents in pure natural language.

Powered by a customized Large Language Model combined with LangChain, the Agent deciphers queries like "Swap 100 USDC for TSLA" or "What is the current market sentiment?". It autonomously prepares transaction calldata, validates it through the Stylus guardrails, and executes it via the Paymaster.

The Intelligence Vault represents the next evolution of yield generation. Rather than relying on static smart contract logic, this vault is actively managed by an autonomous Multi-Agent Consensus System.

• Macro Analysis: Constantly monitors real-time market data (RSI, MACD, Pyth Network Oracles) across crypto and tokenized RWA stocks.
• Automated Execution: Automatically rebalances the portfolio (e.g., swapping aUSD to WETH when bullish momentum is detected) using the backend execution pipeline.

The Pro Trading Dashboard provides an immersive, cyberpunk-themed interface for manual and AI-assisted perpetuals trading.

It features real-time TradingView charts, an interactive order book, and instant execution through the Aura Smart Account. Traders can toggle between Long/Short positions, adjust leverage with zero slippage tolerance, and leverage AI signals directly integrated into the trading view to make informed decisions on both Crypto and Tokenized Stocks.

Aura deploys a Rust/WASM perpetual order book on Arbitrum Stylus — the first hackathon project to combine a WASM LOB with a Solidity Vault LP for hybrid execution.

• 34% gas savings on get_active_orders_sorted vs pure Solidity at scale (60 resting orders). Stylus excels on compute-heavy hot paths.
• AI Market Maker places symmetric bid/ask quotes around Pyth mid price. AI Keeper polls Pyth every 10s and calls match_orders to fill triggered limits.
• Cross-chain hybrid: Stylus LOB on Arbitrum Sepolia (compute layer) + AuraPerps on Robinhood Chain (settlement layer). Market orders hit the Vault LP; limit orders rest in the WASM book.

Every user intent passes through a dual-agent safety architecture before reaching the blockchain:

• Executor Agent (NVIDIA Llama 3.1 70B): translates natural language into precise on-chain calldata — swap routing, DCA scheduling, limit order encoding.
• Risk Auditor Agent: independently verifies balances, allowances, slippage, and macro context (Pyth + NewsAPI + correlation matrix) before approving execution.
• Result: the user sees a single signature prompt with full reasoning visible. If the Auditor rejects, the trade is blocked with an explanation — no silent failures.

The Robinhood Chain is uniquely positioned to onboard the next million retail users into Web3. Aura is built natively for this ecosystem to provide the "missing link": an intelligent, gasless copilot that makes interacting with DeFi and Real World Assets (RWAs) as intuitive as using a traditional fintech app.

With Account Abstraction (EIP-4337), users never need to hold ETH for gas. The AuraPaymaster sponsors transactions, and the AuraAccount smart wallet enables batch execution — multiple approvals and swaps in a single signature.

Aura deploys a second Stylus WASM contract — the AuraGuardrail — as an on-chain safety layer that validates every trade before execution, even if the AI agent is compromised.

• 5 on-chain checks: asset whitelist, max leverage (50x), min collateral, max position size, daily volume cap per user — all enforced in WASM at near-zero gas cost.
• Audit Trail: Before every gasless execution, the agent records a keccak256 hash of its full reasoning (executor + auditor + macro analysis) on-chain via the AuraAuditTrail contract. This creates a verifiable, immutable proof that the AI committee approved the trade.
• Slippage Protection: Every swap calculates a minAmountOut from real-time Pyth oracle prices (1% max slippage). No more zero-slippage exploits.
• Real-time Reasoning Stream:The multi-agent committee's decision process is streamed live to the frontend via SSE — users see each step (Intent Parsing → Macro Audit → On-Chain Audit → Consensus) as it happens.

Aura exposes a Model Context Protocol (MCP) server so any AI agent — Claude, ChatGPT, Cursor, Kiro, or custom agents — can connect and trade perpetuals, swap tokens, manage positions, and analyze markets by simply adding a URL.

• 24 tools exposed: get_price, get_orderbook, place_limit_order, get_limit_orders, place_market_order, get_positions, close_position, get_account_balance, set_stop_loss_take_profit, add_margin, get_market_analysis, get_funding_rate, partial_close, dca_order, cancel_dca, schedule_swap, get_audit_trail, get_pnl_summary, get_liquidation_price, get_supported_assets, cancel_limit_order, swap, deposit_vault, authenticate.
• Delegation model: Users authorize the MCP agent via setAiAgent() on their AuraAccount. No private key is ever exposed. Revocable at any time. Per-user API keys for authenticated trading.
• On-chain audit trail: Every trade executed via MCP automatically records a reasoning hash + AI confidence score (0-100) on-chain in the AuraAuditTrail contract — verifiable proof of AI decision-making.
• Multi-transport: SSE (Claude Code, Cursor) + Streamable HTTP (Claude.ai web, ChatGPT) — compatible with all major AI platforms.
• No other DeFi project has this. The most complete DeFi MCP server ever built. Any AI in the world can read our Stylus order book, execute trades, manage risk, and earn yield via a single URL endpoint.

Aura enforces strict separation of concerns at the private key level to eliminate nonce collisions and secure the autonomous trading loop.

• Dedicated Chat Wallet: The AI Agent MCP and the gasless relayer operate on a separate wallet (the Chat Wallet) authorized via setAiAgent on the AuraAccount. This ensures AI intents never conflict with core protocol execution.
• Router-Level Abstraction: Swaps are executed through the Synthra V3 Router, shielding the Vault LP from direct external calls. The Agent only interacts with the router.