85Score
TradingAgentsTauricResearch/TradingAgentsA multi-agent research desk: analyst agents argue the bull and bear case, a risk agent signs off before any trade idea moves.
★ 118k stars · Intermediate · updated 2026-10
AI Agents deep dive
AgenticTrading Review | A University Lab Where AI Models Trade on Paper
AgenticTradingOpen-Finance-Lab/AgenticTradingAgenticTrading, also called Agentic Trading Lab, is a research playground from 
Open Finance Lab where you build an AI trading agent from a model, a data source and a prompt, then test it on history and on simulated fills. It logs every decision with the agent's reasoning and ranks models on a public leaderboard. The catch: it is an experimental academic project, parts of it are still stubs, and there is almost no community feedback yet.
Think of it as a driving school with a closed track: AI models take laps, every turn of the wheel is recorded, and nobody hands them the keys to the highway until an instructor says so.
FullStack Alpha Editorial Team
Reviewed by Jay Rocco, Chief Editor
What Is AgenticTrading and What Does It Actually Do?
Everyone wants to know whether 
ChatGPT or 
Claude can pick stocks. Few people test it properly. AgenticTrading is a university-lab answer to that question: a platform where large language models (LLMs, the AI systems behind chatbots) act as traders, and every decision gets logged, measured and compared.
The workflow is straightforward on paper. You pick a model, a data source and a trading prompt, or start from a template in the Community tab. You run a backtest (replaying your strategy on old prices to see what would have happened), then move the same agent to live-market paper trading, meaning fake money against real prices. A dashboard shows positions, trades, drawdown (the drop from a peak to a low) and the written reasoning behind each action.
The leaderboard is the most useful public piece. It advances nightly after the US market close, trades each new session on simulated fills with no real money, and compares LLMs against baseline strategies and market indices over the same window. The README says the roster is curated, so you cannot submit your own agent to the public board.
Under the hood, the repo also carries the FinAgent orchestration framework, described in a paper accepted at a 
NeurIPS 2025 workshop on generative AI in finance. It coordinates multiple agents with memory and a planning graph, a map of which agent does what in which order. The project's example settings file lists keys for OpenAI, Anthropic, 
DeepSeek, 
OpenRouter and Alpaca, and there is an off-by-default option to connect a real Robinhood account with per-order risk caps and a review-only mode.
Community Score
62 / 100
- Adoption · How many people starred it
- 17.4 / 30
- Maintenance · How recently the code was updated
- 25 / 25
- Community sentiment · What traders say on Reddit and Hacker News
- 8 / 20
- Builder activity · How many people built their own copy
- 7.7 / 15
- License clarity · Whether you are clearly allowed to use it
- 4 / 10
GitHub data checked 2026. Community read: not enough independent discussion yet.
View Open-Finance-Lab/AgenticTrading on GitHub ↗
Is AgenticTrading Right for You?
Good fit if you are
- Students and researchers studying whether AI models can make sensible trading decisions
- Developers who want a ready-made harness to compare LLMs on the same market window
- Curious traders who want to read an AI's reasoning before trusting any AI stock tool
- People comfortable with Python who want backtests and paper trades without building the plumbing
Skip it if you are
- You want a finished product that trades for you. The project calls itself experimental, and its paper-trading backend is labeled a stub in the code layout
- You do not want to manage several API keys and a Python setup
- You need a clear, standard open-source license. It uses OpenMDW-1.0, a newer license built for AI model materials that many readers will not recognize
The jargon, in plain English
- LLM
- Large language model: the kind of AI behind ChatGPT, Claude and similar chatbots.
- Agent
- An AI set up to take actions on its own, like deciding to buy or sell, instead of only answering questions.
- Backtest
- Replaying your strategy on old prices to see what would have happened.
- Paper trading
- Trading with fake money against real prices.
- Simulated fills
- Pretend order executions. The system records a trade at a market price, but no real order is sent.
- Drawdown
- How far an account falls from its high point before it recovers. A key measure of pain.
- API key
- A password your code uses to talk to a service, like an AI model or a broker.
What Do Traders Say About AgenticTrading?
There is no community verdict to report. Our search of 
Reddit and 
Hacker News found no relevant discussion of AgenticTrading at all. For a research project, that is normal. Academic tools tend to get cited in papers, not debated on r/algotrading.
The GitHub numbers suggest real interest anyway: close to 800 stars, more than 150 forks (a fork is someone's copy they build on) and code pushed the day before we checked. That is a lot of activity for a lab project. The missing piece is independent users saying what worked and what broke.
What they like
- Very active development
- Decision logs show the AI's reasoning for each trade
- Public leaderboard compares models on the same market window
What they complain about
- No relevant community discussion found
- Experimental status with some unfinished parts
How Do You Set Up AgenticTrading Safely?
Plain steps, fake money first. Keep your keys in a private settings file, never pasted into the code.
- 01Look at the hosted dashboard firstBrowse the public dashboard and leaderboard on the project's
Vercel site to see what the tool does before you install anything. - 02Create a Python 3.13 environmentThe README uses uv, a fast Python installer, to make a fresh environment with Python 3.13 so the project's packages do not clash with others on your computer.
- 03Install the requirementsInstall the packages listed in requirements.txt and requirements-vnpy.txt with uv, as shown in the README's Run Locally section.
- 04Run the safe local modeRun dashboard/scripts/run_local_safe.sh and open the local address it prints. The README says this mode needs no keys and nothing can trade or bill.
- 05Add keys in the environment fileCopy .env.example to .env and add only the keys you need, such as one AI model and
Alpaca's paper-trading key. Keep keys there, never in code or notebooks. - 06Backtest, then paper tradeRun a backtest, read the decision logs, then move to paper trading. Leave the
Robinhood live option off unless you have weeks of paper results you understand.
What Do You Need to Run AgenticTrading?
Python and the uv installer
The project runs locally as a Python app with a web dashboard.
An AI model API key
Agents need an LLM to make decisions. The example settings file includes slots for several providers and gateways.
OpenAIAnthropicDeepSeekOpenRouterA broker account for market data and paper trading
The settings file points to Alpaca's paper-trading service by default. A live Robinhood link exists but is off by default.
A host (optional)
Only if you want to deploy your own copy. The repo includes configs for these platforms.
RenderVercel
What Are the Risks of Using AgenticTrading?
Experimental by design
The README calls it an experimental playground, and the code layout marks the paper-trading backend as a stub. Expect rough edges and changes.
LLMs are not proven traders
A model that writes a convincing reason for a trade is not the same as a model with an edge. Read the logs with skepticism, and judge results on drawdown as well as returns.
Real-money option exists
You can connect a real Robinhood account. It is off by default with risk caps, but an AI placing real orders on an experimental platform is a risk most people should not take.
Unfamiliar license
OpenMDW-1.0 is permissive, but it is new and written for AI model materials. Check it before building anything commercial on top.
API costs add up
Every agent decision is a call to a paid AI model. Long backtests across many agents can run up a bill if you do not watch usage.
Keep reading across FullStack Alpha
- AIStockTradingBots.comWhat agentic trading really means →
- AIStockTradingBots.comTradingAgents, our full breakdown →
- AIStockTradingBots.comAI trading agent risks →
- AIStockTradingBots.comBuilding a Claude trading agent →
- AIStockPickerApps.comAI trading platforms directory →
- AIStockPickerApp.comAI stock picker claims, fact-checked →
Not ready to run code yourself? Our hands-on AI stock tool reviews on AIStockPickerApp.com score finished products on five pillars, the AI stock tools directory at AIStockPickerApps.com sorts 200+ tools by the job they do, and AIStockTradingBots.com walks through broker APIs, bot setup and backtesting step by step. Or head back to the Trading Workflows home page to match tools to your own routine and budget.
What Are the Best AgenticTrading Alternatives?
- 85ScoreTradingAgentsTauricResearch/TradingAgents
84Score
Vibe-TradingHKUDS/Vibe-TradingA multi-agent research workspace from HKU with built-in backtest engines and trade-journal analysis.
★ 37k stars · Intermediate · updated 2026-10
- Claude
- Codex
Local models
83Score
AI Hedge Fundvirattt/ai-hedge-fundInvestor-persona agents debate a stock, then a portfolio agent makes the call. A teaching project, not a fund.
★ 64k stars · Intermediate · updated 2026-10
GeminiShould You Use AgenticTrading? Our Verdict
AgenticTrading is one of the more honest AI trading projects on GitHub. It does not promise profits. It builds a lab: backtests, paper trades, decision logs and a leaderboard that puts LLMs next to boring baselines and index funds. That comparison alone teaches more than most AI stock-picking marketing.
Use it as a research tool, not a trading system. If you want to see whether an AI model's stock calls hold up, this lab is a solid place to watch it fail or succeed with fake money. Keep the live broker link switched off. Systems beat hacks, and an unproven AI with your brokerage password is the biggest hack of all.
More AgenticTrading Questions, Answered
What is AgenticTrading?
AgenticTrading, or Agentic Trading Lab, is an open-source research platform from Open Finance Lab for building and testing AI trading agents. You can backtest, paper trade, read each agent's reasoning and compare models on a leaderboard.
Is AgenticTrading free?
The code is free under the OpenMDW-1.0 license. You pay for any AI model API you use, and for hosting if you deploy your own copy.
Which AI models work with AgenticTrading?
The example settings file has slots for OpenAI, Anthropic, DeepSeek and OpenRouter keys, plus a gateway that covers several providers. You pick the model when you create an agent.
Does AgenticTrading trade real money?
It can, through an optional Robinhood connection that is off by default and limited by per-order risk caps. The main workflow is backtests and paper trading with simulated fills.
Can an AI agent beat the market on AgenticTrading?
That is what the leaderboard tests, by comparing AI models with baseline strategies and market indices over the same period. Treat any result as research, not a promise of future returns.