UK Edition 19 November 2026 Independent · Editorial · No paid placements

Lead analysis · AI Analytics

AI Agents in Trading: What They Actually Do — and What They Still Can't

Autonomous agents have moved from conference demos into real research workflows. A sober look at where they add value for traders, where the risks concentrate, and how to evaluate them.

Simulated equityDrawdown
Illustration only: a simulated equity curve and its drawdowns, generated for this page. Not real or expected performance.
Articles
15
Tool profiles
16
Calculators
6
Course lessons
8
Glossary terms
28

In this edition

Editors' picks

All 15 articles →

Latest

Analysis, guides & explainers

Our approach

Reporting on trading technology — without selling it.

  • I.

    We publish no prices, affiliate links or paid placements.

  • II.

    Every profile is sourced from public documentation and dated.

  • III.

    Corrections are logged openly under our Editorial Policy.

Research database

AI Agents & Trading Systems Directory

Structured editorial profiles of 16 tools used for research, analytics and systematic trading.

Browse the full directory →

FinRobot

Agent framework

  • AI Agents & LLMs
  • Open-source

Open-source platform from the AI4Finance Foundation that orchestrates LLM-based agents for financial analysis tasks such as equity research reports and forecasting workflows.

Asset coverage
Equities, multi-asset research
Intended autonomy
Human-in-the-loop
Editorial note

Best viewed as a research scaffold rather than a finished product. Teams with Python experience can adapt its agent roles to their own data; outputs still require analyst review.

Official project / vendor site ↗

Full profile →

FinGPT

Financial LLM project

  • AI Agents & LLMs
  • Open-source

Open-source financial language models and data pipelines focused on lightweight fine-tuning for tasks such as sentiment classification and financial Q&A.

Asset coverage
News, filings, sentiment
Intended autonomy
Human-in-the-loop
Editorial note

Useful for experimentation with domain-adapted language models. Benchmark claims should be reproduced on your own data before being relied on.

Official project / vendor site ↗

Full profile →

OpenBB

Research platform & SDK

  • Market Data & Research
  • Open-source

Open-source investment research platform and Python SDK that aggregates many financial data providers behind a consistent interface.

Asset coverage
Multi-asset
Intended autonomy
Human-in-the-loop
Editorial note

Strong choice for analysts who want scriptable access to many data sources. Data quality and licensing depend on the underlying providers you connect.

Official project / vendor site ↗

Full profile →

Freqtrade

Trading bot framework

  • Algorithmic Execution
  • Open-source

Python crypto trading bot with backtesting, hyperparameter optimisation, dry-run mode and control via web UI or Telegram; exchange connectivity through CCXT.

Asset coverage
Crypto
Intended autonomy
Fully automated execution
Editorial note

Mature community project with thorough documentation. Its optimiser makes overfitting easy, so out-of-sample validation is essential.

Official project / vendor site ↗

Full profile →

Hummingbot

Market-making framework

  • Algorithmic Execution
  • Open-source

Open-source framework for building market-making and arbitrage strategies across centralised and decentralised exchanges.

Asset coverage
Crypto (CEX & DEX)
Intended autonomy
Fully automated execution
Editorial note

Designed for liquidity provision rather than directional trading. Inventory and counterparty risk need explicit controls.

Official project / vendor site ↗

Full profile →

QuantConnect (LEAN)

Algorithmic platform

  • Algorithmic Execution
  • Open-source engine

LEAN is an open-source algorithmic trading engine for C# and Python; QuantConnect provides hosted research, backtesting and live deployment around it.

Asset coverage
Equities, futures, options, FX, crypto
Intended autonomy
Fully automated execution
Editorial note

One of the most complete end-to-end environments available to individuals. The engine can be self-hosted, which reduces platform lock-in.

Official project / vendor site ↗

Full profile →

Interactive

Tools for thinking about risk

Educational calculators that run in your browser. Nothing you enter leaves your device.

All 6 tools →

Free course

Systematic Trading 101

A free, eight-lesson introduction to how systematic trading works, how strategies are tested, and why risk management comes first. No prior experience required.

  1. 01What Is Systematic Trading?
  2. 02From Idea to Testable Hypothesis
  3. 03Data Foundations
  4. 04Backtesting Basics
  5. 05Costs, Slippage and Execution
View all 8 lessons →

Side by side

Comparisons

All comparisons →

Coverage

Browse by discipline

Reference

From the Glossary

Agentic AI
Software that pursues a goal by planning steps, calling tools or data sources, and adjusting based on results, rather than answering a single prompt.
Alpha
Return in excess of a benchmark after adjusting for the risk taken. Persistent alpha is rare and tends to decay as more participants discover the same edge.
Backtest
A simulation of how a trading rule would have performed on historical data. Useful for falsifying ideas; weak evidence of future returns on its own.
Conditional VaR (CVaR)
Also called Expected Shortfall: the average loss in the worst cases beyond the VaR threshold. Captures tail severity that VaR ignores.
Full glossary →