
How To Build a Complete AI Trading Bot In Under 20 Mins (full guide)
This guide shows how to build an AI-powered crypto trading bot from scratch using Claude (or ChatGPT) and Bybit’s native MCP integration. No coding required — the bot handles execution, logging, and learning automatically.
Prerequisites- Paid subscription to Claude (or ChatGPT Codex) — at least the $20 plan.
- Node.js installed on your machine (the setup prompt handles this).
- Exchange account with native MCP support (Bybit recommended due to official integration and single-command setup).
- Create an exchange account and enable 2FA.
- Generate an API key with trading permission (no withdrawal access).
- Connect the MCP to Claude using a single command, then restart Claude.
- Test the connection by asking for your balance or placing a small test order.
- Set up a memory system – a folder containing:
context.md(your financial situation, goals, strategy)instructions.md(how the bot should behave)memory.md(automatically updated after each trade)
- Ledger – logs every trade: when it was opened, why, and the outcome.
- Learnings file – the bot writes plain‑English lessons after each closed trade.
- Forced reflection – before opening a trade, you add your justification and invalidation criteria. This helps improve future decisions.
- Create a Telegram bot via BotFather and connect it to your always‑on Claude instance (e.g., on a Mac Mini or VPS).
- Use voice messages (transcribed by Whisper Flow) to check portfolio, place trades, and receive alerts while mobile.
- Use algorithmic strategies (e.g., moving average crossover) backtested in a separate app.
- Or discretionary rules (e.g., “buy Bitcoin when it drops 10% with IV > 1.5”).
- The bot is an execution layer — you can plug in any strategy and the memory system refines it over time.
- Not a “set and forget” money printer – it requires strategy tuning and manual review of accumulated lessons.
- Errors can happen (MCP connection, API issues) – use Claude itself to diagnose and fix them.
- Start with a sub‑account or paper trading to avoid risking main funds.
- Over time, compound your trade data to build a systematic edge before most traders adopt AI.
- Live experiments with real money.
- Stock trading via Interactive Brokers.
- Deeper dives into Telegram and custom indicators.






