LLM-Directed Crypto Perpetual Futures Trading with Automated Risk Controls

Summary

This document outlines an automated system that uses a large language model to produce entry, hold, or close decisions for multiple cryptocurrency perpetual futures. Its pipeline collects market and account data, calculates indicators across short and long timeframes, combines them with open-position details, and sends the context to an AI decision engine. A separate execution layer sizes orders from the proposed risk and stop distance, applies exchange precision rules, and monitors take-profit and stop-loss levels.

The described safeguards include position limits, per-trade risk constraints, minimum risk-reward requirements, explicit invalidation conditions, and a rule against adding to an existing position. The workflow source indicates a three-minute schedule, while the overview describes minute-level operation; this cadence is not consistent. The document provides design details, not a measured trading evaluation: it reports no backtest or live performance results. AI-generated decisions, leveraged exposure, execution differences, and the reliability of the prompt and monitoring logic remain important limitations.

Key ideas

• The system separates market perception, AI decision-making, and trade execution into a recurring workflow.
• It supplies the model with indicator data from two timeframes and current account and position state.
• The execution layer derives position size from risk amount and stop distance and monitors exit levels.
• Position limits, stop losses, profit targets, and invalidation rules are intended to constrain risk.
• The document describes architecture but provides no performance evidence, and its stated operating cadence differs from the workflow schedule.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.