Descriptions:
Michael Automates explores OpenClaw, an agentic AI platform formerly known as Claudebot, that can autonomously write, backtest, schedule, and execute cryptocurrency trading strategies without human intervention between steps. Unlike standard AI chatbots such as Claude or ChatGPT, OpenClaw can install Python and JavaScript packages, write and run code, set up Linux cron jobs for timed execution, and interact directly with exchange APIs — demonstrated here against a live HyperLiquid account.
The hands-on walkthrough shows OpenClaw writing separate strategies for trending and ranging market conditions in JavaScript, running backtests on historical data, selecting the optimal approach, and then autonomously executing trades. Actual trade logs are shown, including long and short Bitcoin positions with real profit and loss figures, and the creator demonstrates running multiple strategies in parallel on different timeframes.
A substantial portion of the video addresses the security and infrastructure reality of running an agentic system: OpenClaw inherits full system-level permissions from whatever machine it runs on, creating risks around file access or deletion. The creator evaluates practical hosting options including dedicated Mac Mini setups and cloud virtual machines via DigitalOcean and Emergent.sh, concluding that most beginner-friendly solutions fail for trading because the machine suspends between sessions, breaking scheduled strategy execution.
📺 Source: Michael Automates · Published February 04, 2026
🏷️ Format: Hands On Build
