
how to actually build a 24/7 ai trading agent (full guide)
The system consists of an always-on machine (e.g., Mac Mini, old desktop, or $10/month VPS) running Claude Code 24/7. It connects to:
- Exchange layer: Bybit (crypto) and Alpaca (equities) with paper trading support
- Data connectors: FMP (live stock/crypto data) and Quiver (insider trades, congressional filings)
- Memory folder: Contains your trading strategy, trade ledger, personal investor context
- User interface: Telegram bot (via BotFather) for voice/ text commands
- Always-on device → Install Claude Code (or Codeex)
- Exchange API → Generate API keys (disable withdrawals, enable 2FA)
- Connectors → Add FMP and Quiver MCPs in Claude settings
- Memory folder → Store your strategy, trade history, and personal info
- Telegram bot → Create via BotFather, paste token into Claude Code
- Voice input (optional) → Use Whisper Flow to convert speech to text
- Portfolio analysis: "Break down my portfolio by sector" → instant breakdown
- Alpha alerts: "Notify me when a semiconductor company has major insider accumulation" → automated Telegram alerts
- Trade execution: "Buy $100 worth of Nvidia stock" → order placed on brokerage
- Scheduled orders: "Set an order for tomorrow at 9 AM to rebalance 5% into gold and T-bills"
- Start with paper trading (Alpaca default, Bybit testnet)
- Limit exposure: Only connect a sub-portfolio, not full net worth
- Security: Disable API withdrawals, enable 2FA, harden setup over time
- Early stage: This is version 1; expect consumer-grade products in ~3 years
You can now control your trades from anywhere via voice commands on your phone. The system autonomously handles market insights, alerts, and execution. Join the free school community (link in description) for the one‑prompt setup guide and all resources.






