AlphaTrading

AlphaTrading is a multi-agent AI trading framework designed to bridge advanced off-chain quantitative analysis with verifiable on-chain execution. Built from first principles, it moves beyond simple single-prompt wrappers by introducing a collaborative "AI Syndicate" architecture. The system leverages LangGraph to orchestrate a multi-stage workflow. Specialized AI analysts (Market, News, and Quant) process cross-domain data signals, which feed into a rigorous Bull/Bear research debate. This ensures every strategy is deeply stress-tested before a Trader agent generates a preliminary TradeIntent. A critical technical differentiator of this framework is the strict decoupling of AI reasoning from financial execution. To ensure capital safety, a deterministic Risk Engine intercepts all AI-generated proposals. It enforces hard constraints—such as maximum position sizing and single-order limits—before any action is finalized. Furthermore, the system maintains a continuous learning loop, utilizing ChromaDB for semantic memory and SQLite for robust portfolio state tracking. To solve the "trust black box" inherent in autonomous AI, the framework natively integrates with the ERC-8004 standard. The agent registers a verifiable on-chain identity, signs trade intents using EIP-712, and interacts with a smart contract RiskRouter. By waiting for asynchronous on-chain feedback to update its virtual ledger, the agent creates a transparent, immutable track record, building a measurable reputation based entirely on objective, risk-adjusted performance.