LLM Council Expert Trader

LLM Council Expert Trader is an Expert Advisor that uses Large Language Models to analyze market structure and manage trades based on Smart Money Concepts (SMC) and Inner Circle Trader (ICT) methodology. Rather than relying on static indicator signals alone, the EA sends market context to AI models for structured reasoning before any trade is placed.

LIVE TRADING PERFORMANCE SHOWROOM OF THIS EXPERT ADVISOR AVAILABLE AT https://www.mql5.com/en/signals/2383309?source=Site+Profile#!tab=stats
Disclaimer: Results depend on market conditions, model selection, and EA settings. Past performance does not guarantee future results

Multi-Agent Architecture

The EA operates through four specialized AI agents that collaborate on each trading decision:

• HTF Agent: Analyzes higher-timeframe bias (D1/H4/H1) using SMC concepts, liquidity levels, and AMD phases.
• Structure Agent: Identifies Market Structure Shifts, Break of Structure, Fair Value Gaps, and Order Blocks on M5.
• Strategy Agent: Selects or adapts strategies based on current volatility and recent performance.
• Execution Agent: Validates risk parameters, including minimum R:R of 1:2, spread checks, and portfolio exposure.

Supported AI Providers

The EA supports multiple LLM providers with per-agent routing:

• OpenRouter API: Access to cloud models such as Claude, GPT, and Qwen.
• Together.ai: Cloud inference with support for custom LoRA adapters and fine-tuned models.
• LM Studio: Local inference using GGUF-compatible open-source models at no API cost.
• Ollama / llama.cpp: Compatible with self-hosted or cloud-based inference (Vast.AI, DigitalOcean, etc.).

Hybrid routing allows mixing local and cloud models across agents. The priority order (LM Studio, Together.ai, OpenRouter) is fully configurable with automatic fallback.

Local Inference with LM Studio

The EA can run entirely on your own hardware using any GGUF-compatible model. Suggested models for local use:

• Qwen2.5-14B-Instruct-GGUF: Balanced reasoning quality and VRAM usage.
• Meta-Llama-3.1-8B-Instruct-GGUF: Lightweight, approximately 8 GB VRAM.
• Qwen2.5-72B-Instruct-GGUF: Strong reasoning quality, requires approximately 48 GB VRAM.
• DeepSeek-R1-Distill-Qwen-14B: Suitable for structured output in the Execution Agent role.

Custom Model Support

Traders and developers can deploy their own LoRA adapters, fine-tuned models, or custom structure agents. Different models can be assigned per agent role (HTF, Structure, Strategy, Execution), using cloud or local inference.

Risk and Position Management

The EA includes built-in risk controls:

• Dynamic position sizing from 0.25x to 2x base risk.
• Automatic risk reduction when drawdown exceeds 5%.
• Breakeven and trailing stops placed behind market structures.
• Partial profit-taking at configurable levels (default: 30% and 60% of target).
• Auto-close on adverse structure change or trade inactivity.

Setup Instructions

1. Set your OpenRouter API key (optional if using LM Studio only).
2. Download the required indicators (.ex5) from the Comments section and place them in the MQL5\Indicators folder.
3. Attach the indicators and then the EA to M5–H1 charts.
4. Monitor the Experts log for agent decisions and trade activity.

For LM Studio setup: install LM Studio, download a GGUF model, enable CORS, and start the local server on port 1234. In MT5, add http://localhost:1234 to the WebRequest allowed list (Tools > Options > Expert Advisors). In EA inputs, set InpUseLMStudio to true.

Add these URLs to the WebRequest allowed list in MT5: https://openrouter.ai and https://api.together.xyz. Add http://localhost:1234 if using local inference.

Configuration presets for Forex and Crypto are available in the Comments section.

Summer Holidays Promotion (until August 19, 2026)

1-Month EA Rental: $43.99 (regular price: $100).
Lifetime EA License: $443.99 (regular price: $5,000). No recurring EA fees. Only API usage costs apply, or run at no cost with LM Studio local inference.

Important Notice

This Expert Advisor is a technical tool, not investment advice. It does not constitute a recommendation to buy, sell, or hold any financial instrument. Trading foreign exchange, CFDs, and other leveraged products involves significant risk of loss and is not suitable for all investors. You should carefully consider your financial situation and risk tolerance before trading. Never trade with money you cannot afford to lose. Results depend on market conditions, model selection, and EA settings. Past performance does not guarantee future results. Always validate new models or configurations on a demo account before live deployment. Estimated cloud API cost per analysis is approximately $0.01–$0.05; local inference has no API cost.

All input parameter names, and output messages are in English.