BARCO-Alpaca AI Trading Agent

BARCO-Alpaca is an autonomous AI trading agent that uses Claude claude-opus-4-5 as its decision brain, connected to Alpaca's paper trading API through MCP (Model Context Protocol) tools. SIGNAL INTELLIGENCE STACK The agent receives real-time intelligence from three independent production systems before every decision: • VecFrachZ — quantum-validated directional signals for BTC, ETH, SOL, ADA and more. Signals are processed using IBM Quantum circuits (Theta=72°, Alpha=2.1875) achieving 0% bit error rate. This gives the agent a unique edge in crypto signal quality. • S2 OrderBook CNN — a convolutional neural network that reads live order book snapshots every second and outputs a buy/sell pressure score from 0.0 to 1.0. Scores above 0.65 indicate strong buy pressure. • S3 NLP Sentiment — combines the Fear & Greed index with live financial news headlines to produce a market sentiment score per symbol. All three signals are combined by Meta-Brain, an aggregator that produces a single score from -1 (AVOID) to +1 (OPTIMAL). HOW CLAUDE DECIDES Every 15 minutes, Claude receives the full intelligence report and calls MCP tools: get_account, get_positions, get_latest_quote, and get_options_chain. Claude only trades when Meta-Brain is OPTIMAL (score above +0.3) — neutral or negative markets mean HOLD, no exceptions. Maximum $2,000 per trade, maximum 5 open positions. AUTOMATIC RISK MANAGEMENT A Python monitor runs every 30 seconds independently of Claude. It closes any position automatically at +2% take-profit or -3% stop-loss, locking in gains and cutting losses without waiting for the next Claude cycle. LIVE RESULTS Since deployment the agent has opened positions in NVDA, BTC/USD, ETH/USD, and SOL/USD. Current portfolio: $100,013 on a $100,000 starting balance, with all four positions in profit.