🏛️ FinRL-X-MT5

K-Dense Council: Multi-Agent Mixture-of-Experts Trading System for MetaTrader 5

🌐 Official Production Platform & Monetization Portal: primeclub-quant.vercel.app
📖 Official Technical Wiki & Architecture Guide: github.com/ElMoorish/FinRL-X-MT5/wiki
Access real-time multi-agent deliberations, interactive prop challenge drawdown calculators, daily institutional recaps, and pre-trained production model checkpoints.

FinRL-X-MT5 is an institutional-grade algorithmic trading framework that adapts the Deep Reinforcement Learning principles of FinRL to MetaTrader 5.

It eliminates standard single-agent failure modes through the K-Dense Council: a 5-expert Mixture-of-Experts (MoE) architecture with NSGA-III Pareto-optimal gating, fusing high-frequency order-book microstructure with cross-asset equity/macro fundamental flows under the 100% Real Ticks backtesting standard.

---

🖥️ Live Council Terminal & Dashboard

FinRL-X-MT5 features a high-performance local web dashboard with real-time TradingView Lightweight Charts, a live 5-expert deliberation room, prop firm guardian circuit breaker gauges, and an autonomous step-by-step Chain-of-Thought (CoT) reasoning stream.

Dark Mode: Real-time Multi-Agent Deliberations, Prop Firm Guardian Gauges & Step-by-Step CoT Stream

White Mode: Clean Institutional Layout with Precision Tick Feed & Solvency Meters

---

📚 Quantitative Research & Engineering Whitepapers

Explore our comprehensive technical whitepapers covering machine learning topology, actuarial risk bounds, and regime classification:

• 📑 Multi-Agent Reinforcement Learning for MetaTrader 5: Overcoming single-agent non-stationarity via the 5-Specialist Council and NSGA-III Pareto gating.
• 🛡️ Defensive Capital Allocation & 0.50% Drawdown Guards: Mathematical position sizing, circuit breakers, and surviving prop firm evaluation limits.
• 📈 Hidden Markov Models (HMM) for Regime Detection on NAS100: Gaussian HMM state decoding, Baum-Welch training, and filtering choppy consolidation.

---

📖 Official Technical Wiki & Knowledge Base

Explore our comprehensive, production-grade technical documentation hosted on the GitHub Wiki:

• 🏛️ Wiki Home: System Blueprint & Topology
• 🧠 01. The 5-Agent Council Architecture — Deep dive into SAC, HMM, TimesFM, XGBoost, and PyMC Actuary.
• 🛡️ 02. Risk Management Architecture & 0.50% Capital Ceiling — Sizing formulas, contract multipliers, and Rule 8.
• 🎯 03. The +1.0R Dynamic Breakeven Engine — Trailing stop mechanics and spread buffer defenses.
• ⚡ 04. MetaTrader 5 Bridge Setup & Live Deployment Guide — Terminal flags, .env credentials, and daemon execution.
• 🏆 05. The Prop Firm Passkeeper Evaluation Protocol — 3-stage circuit breakers for FTMO and Goat Funded Trader.
• 🖥️ 06. Local Monitoring Dashboard & Telemetry Server — FastAPI endpoints, WebSocket feeds, and ticket audits.

---

🏛️ Architecture Overview

                               ┌──────────────────────────────────────────────┐
                               │             MT5 High-Frequency Ticks         │
                               │          + Yahoo/Macro Correlation Baskets   │
                               └──────────────────────┬───────────────────────┘
                                                      │
                                           ┌──────────▼──────────┐
                                           │  Correlation Fuser  │
                                           │  (Polars M5 Matrix) │
                                           └──────────┬──────────┘
                                                      │
                       ┌──────────────────────────────┼──────────────────────────────┐
                       │                              │                              │
             ┌─────────▼─────────┐          ┌─────────▼─────────┐          ┌─────────▼─────────┐
             │    Expert 2       │          │    Expert 4       │          │    Expert 3       │
             │   HMM Regime      │          │  SHAP + XGBoost   │          │ TimesFM / EWMA    │
             │ (Market State)    │          │  (Signal Scorer)  │          │(Volatility Bands) │
             └─────────┬─────────┘          └─────────┬─────────┘          └─────────┬─────────┘
                       │                              │                              │
             ┌─────────▼─────────┐                    │                    ┌─────────▼─────────┐
             │    Expert 1       │                    │                    │    Expert 5       │
             │    SAC DRL        │                    │                    │ Bayesian Actuary  │
             │ (Action / Conv)   │                    │                    │ (Dynamic TP / SL) │
             └─────────┬─────────┘                    │                    └─────────┬─────────┘
                       │                              │                              │
                       └──────────────────────────────┼──────────────────────────────┘
                                                      │
                                           ┌──────────▼──────────┐
                                           │   NSGA-III Gate     │
                                           │ (Pareto α Weighting)│
                                           └──────────┬──────────┘
                                                      │
                                           ┌──────────▼──────────┐
                                           │   Consensus Signal  │
                                           │  (Direction & Size) │
                                           └──────────┬──────────┘
                                                      │
                                 ┌────────────────────┴────────────────────┐
                                 │                                         │
                       ┌─────────▼─────────┐                     ┌─────────▼─────────┐
                       │  Strategy Tester  │                     │    Live Socket    │
                       │ (Real-Tick CSV)   │                     │    Order Engine   │
                       └───────────────────┘                     └───────────────────┘

The 5 Council Specialists

Specialist	Engine / Algorithm	Institutional Role	Mathematical Basis
E1 — The Trader	Soft Actor-Critic (SAC)	Directional weight & action conviction	Max-Entropy DRL with entropy regularization
E2 — The Oracle	Gaussian HMM (3-State)	Regime detection & position throttling	Bull / Bear / Consolidation hidden states
E3 — The Prophet	Google TimesFM 2.5 / EWMA	60-min forward volatility envelope	Zero-shot deep forecasting & quantile projections
E4 — The Analyst	SHAP + XGBoost	Feature attribution & quality filter	Game-theoretic Shapley values on factor flows
E5 — The Actuary	Bayesian / Student's t	Dynamic Take-Profit & Stop-Loss levels	Posterior predictive credible intervals ($\mathbb{E}[\text{RR}] \ge 1.5$)
The Gate	NSGA-III Genetic Algorithm	Optimal multi-objective expert blending	Pareto frontier: $\max \text{Sharpe}$, $\min \text{DD}$, $\min \text{Turnover}$

---

🌐 Supported Multi-Asset Universe

The framework supports multiple asset classes with automatic contract-size normalization and cross-asset correlation baskets:

• Commodities: WTI.x (Crude Oil), XAGUSD.x (Silver Spot)
• Indices: NAS100.x, US30.x, SPX500.x, GER40.x, JAP225.x, UK100.x, AUS200.x
• US Equities: AAPL.x, NVDA.x, MSFT.x, AMZN.x, META.x, TSLA.x, PLTR.x

---

🚀 Quick Start

1. Prerequisites

• Operating System: Windows 10/11 (Required for MetaTrader 5 Python IPC)
• MetaTrader 5 Desktop Terminal (Build 4000+)
• Python: 3.10 or 3.11 (64-bit)
• CUDA-compatible GPU (Recommended for DRL acceleration)

2. Installation

# Clone repository skeleton
git clone https://github.com/ElMoorish/FinRL-X-MT5.git
cd FinRL-X-MT5

# Install dependencies
pip install -r requirements_mt5.txt

3. Environment Configuration

Copy .env.example to .env:

cp .env.example .env

(Note: If your MT5 terminal is already open and logged in, python attaches automatically without needing credentials in .env)

4. Train the Council

Train a single instrument:

$env:PYTHONPATH="."
python -m src.main_mt5 train --symbol NAS100.x --days 60 --timesteps 20000

Or train a batch of instruments across multiple asset classes:

python -m src.main_mt5 train --symbols WTI.x XAGUSD.x US30.x GER40.x AAPL.x NVDA.x --days 60 --timesteps 15000

5. Export Strategy Tester Signals

Generate aligned signals and automatically copy them to MT5's MQL5\Files\finrl_x_mt5\ directory:

python -m src.main_mt5 export-signals --symbol NAS100.x --days 60

6. Run Real-Tick Backtest in MetaTrader 5

1. Open MT5 and press Ctrl + R to open the Strategy Tester.
2. Select Expert: FinRL_X_MT5.
3. Select Symbol: NAS100.x, Timeframe: M5.
4. Set Model to: Every tick based on real ticks.
5. Click Start to run the backtest.

---

📚 Technical Documentation

Deep-dive documentation is available in the docs/ folder:

• System Architecture — Mathematical formulation of all 5 experts and NSGA-III gate.
• Data Pipeline & Fusion — Microstructure engineering and 1-day lagged fundamental cross-asset edge.
• MT5 Strategy Tester Guide — Step-by-step institutional real-tick backtesting protocol.
• Execution & Risk Management — Half-Kelly sizing, contract size normalization, and circuit breakers.
• Strategy Optimization Roadmap — Empirical levers to maximize Win Rate, Accuracy, and Profit Factor.

---

🔒 Open Source & Privacy Policy

This repository skeleton contains NO private broker account numbers, passwords, server IPs, or proprietary client files.

• .gitignore strictly excludes .env, models/, logs/, SQLite databases, tick caches, and local MT5 data paths.
• All credentials are abstracted via environment variables (pydantic-settings).

---

💎 Commercial Tiers & VIP Alpha Signals

For live execution signals, prop challenge passkeeper rules, pre-calibrated model weights, and institutional licensing, visit the official FinRL-X Prime Quant Portal:

• Tier 1: Prime VIP Alpha ($79/mo): High-conviction Council signals dispatched directly to private Telegram & Discord feeds with automated daily trade recaps and weekly macro tear-sheets.
• Tier 2: Prop Passkeeper ($199/mo): Strict 0.50% actuary lot sizing, pre-market New York bias reports, and challenge preservation governance.
• Tier 3: Quant Pro Model Weights ($1,997): Direct checkpoint weights (SAC, Gaussian HMM, TimesFM, XGBoost, and Bayesian VaR) dropping right into the framework's weights/ directory.
• Tier 4: Enterprise Bespoke ($9,977): Perpetual commercial license, multi-asset checkpoints, and rolling retraining pipeline code.

👉 Explore All Commercial Tiers & Model Checkpoints

---

☕ Support & Donations

If FinRL-X-MT5 has assisted your quantitative research, automated trading operations, or prop firm challenge evaluations, consider buying us a coffee or donating to support ongoing open-source R&D!

Network	Asset / Token	Deposit Address
TRON (TRC20)	USDT (Tether USD)	TC8TFkemSFGEeBPF5ZQKbmjK97FVEGwrwc

TRC20 Deposit Address:
TC8TFkemSFGEeBPF5ZQKbmjK97FVEGwrwc

Note

Network Notice: Please verify that you select the TRON (TRC20) network when sending USDT transfers. Every contribution directly funds GPU compute clusters for model training, tick data storage, and future reinforcement learning upgrades.

---

🙏 Acknowledgements & Lineage

FinRL-X-MT5 is inspired by and builds upon the pioneering Deep Reinforcement Learning foundations of FinRL and FinRL-Trading developed by the AI4Finance Foundation.

While preserving the core financial reinforcement learning principles of the FinRL paradigm, this codebase introduces:

• MetaTrader 5 Real-Time Bridge: Seamless IPC connection to live MT5 terminals and Strategy Tester execution.
• 100% Real-Tick Backtest Protocol: Eliminates bar-interpolation artifacts by aligning Council signals to broker tick streams.
• The K-Dense Council (MoE): Softmax-weighted multi-expert gating (SAC DRL + 3-state HMM + TimesFM/EWMA + SHAP XGBoost + Bayesian Actuary) optimized via NSGA-III Pareto frontiers.
• Microstructure & Cross-Asset Fusion: High-frequency tick metrics combined with lagged ETF and macro factor baskets.

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.