Agentic AI Trading

What if trading could watch, filter, forecast, size, execute, protect, and review every trade under rules you define?

Agentic AI Trading is a practical guide to building guardrailed AI trading agents for modern markets. Instead of asking a chatbot, "What should I buy?" this book teaches you how to design an agentic trading desk: a coordinated system of specialized agents that scan markets, validate data, classify regimes, forecast price paths, score setups, calculate risk, manage orders, control exposure, handle exits, and review performance.
This is not blind automation. It is disciplined autonomy. The trader remains the architect. The agents become the workforce.
Farrukh "Johnny" Malik shows how agentic traders think in loops, guardrails, permissions, state machines, and risk controls. Each chapter introduces a new agent, model, workflow, market class, or production problem, moving the reader from AI-assisted trading to autonomous trading.
What You Will Build
Learn how to organize a full agentic trading desk with watchlist, data quality, regime, forecast, setup scoring, risk, entry, order flow, exit, event, portfolio, review, supervisor, and meta-routing agents.
You will learn how to translate forecasts into trade instructions, block bad data, size positions with discipline, prevent duplicate orders, reduce correlated exposure, manage open positions, and review every decision after the trade closes.
What Each Part Covers
Part I, The Agentic Trading Mindset: Learn why agentic trading is not the same as asking AI for trade ideas, then build the core loop, supervisor logic, forecast-to-instruction process, control plane, and safety guardrails.
Part II, Building the Core Agent Workforce: Create the agents that do the daily work of the desk: scanning, data validation, regime classification, forecasting, setup scoring, risk approval, entry, order flow, exits, events, exposure control, and trade review.
Part III, Autonomous Trading Desks by Market Class: Apply the agentic desk to stocks, ETFs, options, index futures, commodities, forex, and crypto, with different rules for each market.
Part IV, Deployment, Scaling, and the Complete Desk: Assemble the full desk, move through paper trading, micro-sizing, and scaling, and learn how orchestration becomes the final edge.
Models and Methods Covered
The book introduces practical uses of ReAct agents, state graphs, Chronos-2, XGBoost, Isolation Forest, Hidden Markov Models, TimesFM, LightGBM, Quantile Risk Models, CatBoost, Prophet, graph correlation, options pricing, probabilistic forecasting, ensembles, and meta-agent routing.
For Traders and Builders
Non-technical traders can follow the trader path: agent roles, trading workflow, decision logic, risk rules, and desk architecture. Developers, quants, and AI builders can use the optional Advanced AI Concepts: Optional Technical Build sections, which explain prompts, schemas, Python logic, workflows, APIs, and production guardrails.
A bonus Python chapter gives beginners a practical foundation in trading data, moving averages, ATR, watchlist scanning, position sizing, structured agent messages, trade logs, and a mini agentic trading loop.
Also Included
End-of-book indices cover agents, models, markets, risk controls, orders, forecasts, events, Python code, prompts, and troubleshooting. Templates cover agent design, supervisor permissions, data quality, forecast output, setup scoring, risk sizing, order flow, exits, event lockouts, portfolio heat, review logs, and deployment readiness.
The goal is not to remove judgment from trading. The goal is to remove inconsistency, fatigue, hesitation, overconfidence, and emotional interference from repeatable decisions.