📊 Full opportunity report: Introducing Forezai · TradingAgents — a committee of LLMs decides paper-trades on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

TL;DR

Forezai has launched TradingAgents, a system where multiple LLMs collaborate to generate paper-trading decisions. This development aims to explore AI’s potential in financial decision-making without risking real money.

Forezai has launched TradingAgents, a system where a committee of large language models (LLMs) collaboratively decide on paper trades. This development aims to explore whether AI agents can produce trading decisions comparable to human judgment without risking real capital, marking a significant step in AI-driven financial research.

The TradingAgents framework is a fork of an existing multi-agent research platform that structures LLMs into specialized roles, including analysts, debate agents, risk assessors, and decision synthesizers. The system processes structured market data, generates reports, debates opposing theses, and produces trading proposals, all without promising accuracy or predicting markets.

Forezai’s version adds operational layers: an autonomous scheduler, paper-trading interfaces, position management, and a web dashboard. It supports multiple modes, including local simulation, Alpaca paper trading, and a shadow mode that compares simulated trades against potential live execution, with strict safeguards against real trading risks.

The project emphasizes transparency and explicit reasoning, requiring agents to articulate their decisions and the rationale behind them, thus avoiding reliance on the raw context window of the LLMs. The system is designed solely for research, not for live trading or financial advice.

Introducing Forezai · TradingAgents — Thorsten Meyer AI

AGENTS

● ANNOUNCEMENT / MAY 2026

THORSTEN MEYER AI · FOREZAI · § 03

FOREZAI · 03

TRADINGAGENTS · LAUNCH

Research Series · Companion to Polybot Week 1-2 · 2026-05-17

Introducing Forezai · TradingAgents.
A committee of LLMs
decides paper-trades.

After two weeks of finding out most parametric strategies don’t work, the obvious next research question: can multi-agent LLM judgment do any better?

A fork of the open-source TradingAgents framework (TauricResearch): thirteen LLM agents in four stages — four parallel analysts · a bull-bear debate with research-manager arbitration · a three-voice risk team · a two-layer trader + portfolio-manager decision. The fork keeps the agent graph intact and adds the operational layer the upstream doesn’t ship: an autonomous loop · a multi-broker abstraction · a local web dashboard · Codex OAuth · MCP plug-ins · 520+ unit tests. The question is narrower than “do LLMs predict the market” — that prior is “no, with high confidence.” The narrower question is: when LLMs are structured into specialised adversarial roles, does the committee produce decisions at least no worse than a coin flip after fees? Honest priors before running: it might fail too. If it appears to work, the most likely explanation is variance.

Thorsten Meyer AI Forezai · TradingAgents Apache-2.0 fork · upstream cited Companion piece · ~2,500 words Polybot · § 03

This is not financial advice. Nothing in this announcement should be used to inform real trading decisions. The software described trades simulated money by default. If you reconfigure it to trade real money, you should expect to lose that money — regardless of how clever any individual agent’s reasoning looks. Algorithmic trading is zero-sum after fees and structurally hostile to part-time retail strategies.

13 agents

Specialised roles in four stages
Analysts · Debate · Risk · Decision

78% / -33%

Polybot prior: fleet win rate
combined with -33% bankroll

520+

Passing unit tests across engine,
services, HTTP routes (starting baseline)

€0 floor

LLM cost on Codex OAuth
(falls back to public API per token)

FOREZAI / TRADINGAGENTS· APACHE 2.0 FORK· UPSTREAM TAURIC RESEARCH· LANGGRAPH· 13 AGENTS / 4 STAGES· 4 PARALLEL ANALYSTS· BULL-BEAR DEBATE· 3-VOICE RISK TEAM· TRADER + PORTFOLIO MANAGER· 5-TIER FINAL RATING· ALPACA PAPER + LOCAL + SHADOW· LIVE ENDPOINTS HARD-REFUSED· FASTAPI + REACT VIA CDN· CODEX OAUTH· MCP PLUG-IN REGISTRY· 520+ UNIT TESTS· POLYBOT WEEK 1: 21 EXPERIMENTS· WEEK 2: -33% BANKROLL· 78% FLEET WIN RATE· HONEST RESEARCH, NOT EDGE· FOREZAI / TRADINGAGENTS· APACHE 2.0 FORK· UPSTREAM TAURIC RESEARCH· LANGGRAPH· 13 AGENTS / 4 STAGES· 4 PARALLEL ANALYSTS· BULL-BEAR DEBATE· 3-VOICE RISK TEAM· TRADER + PORTFOLIO MANAGER· 5-TIER FINAL RATING· ALPACA PAPER + LOCAL + SHADOW· LIVE ENDPOINTS HARD-REFUSED· FASTAPI + REACT VIA CDN· CODEX OAUTH· MCP PLUG-IN REGISTRY· 520+ UNIT TESTS· POLYBOT WEEK 1: 21 EXPERIMENTS· WEEK 2: -33% BANKROLL· 78% FLEET WIN RATE· HONEST RESEARCH, NOT EDGE·

FIG. 01 — THE 13-AGENT COMMITTEE

Thirteen specialised roles · four stages · biases made to argue in public

The architecture forces the system to articulate its reasoning rather than relying on what a single context window happens to recall

Stage 1 · Four analysts in parallel4 agents

Market

Structure, ranges, regime indicators

News + Insider

News flow, filings, insider activity

Fundamentals

Balance sheet, earnings, ratios

Social Sentiment

Social-media tone, retail signal

↓

Stage 2 · Bull-bear debate + research-manager arbitration3 agents

Bull researcher

Argues upside thesis from analyst reports

Bear researcher

Argues downside thesis from same reports

Research manager

Arbitrates · writes single synthesis

↓

Stage 3 · Three-voice risk team3 agents

Aggressive

Looks for upside · accepts variance

Conservative

Looks for downside · protects capital

Neutral

Balances · forces downside articulation

↓

Stage 4 · Two-layer decision2 agents

Trader

Three-tier proposal · buy / hold / sell

Portfolio manager

Five-tier rating + price target + horizon · sees arguments only, never raw data

The portfolio manager only sees the arguments, never the raw data — which forces the committee to make its reasoning explicit rather than relying on a single context window’s recall. The upstream framework ships the agent graph; it does not ship the operational machinery to run that graph on autopilot, observe its results honestly, store them for later inspection, or prevent the operator from accidentally trading real money. That gap is what the Forezai fork fills.

FIG. 02 — THE POLYBOT PRIOR · WHY THIS IS A DIFFERENT BET

Two weeks of paper-trading prediction markets · the trap underneath the headline numbers

25 experiments · 78% fleet-wide win rate · -33% bankroll · most parametric strategies are structurally negative-expectation when measured honestly

The flattering number

78%

Fleet-wide win rate · week 2

“You can win four out of five trades and still go broke, because the one loss is bigger than the four wins put together.” Win rate without P&L context is a mechanical illusion.

The honest number

−33%

Fleet bankroll · week 2 close

The strongest possible demonstration of the trap. A parametric trading strategy that looks compelling in a backtest will almost always fail to survive a fresh sample. Most “edges” are mechanical artefacts.

Week 1: 21 parallel strategy experiments · early winners mostly mechanical illusions · exactly one strategy (a fair-value taker on BTC) showed the mathematical signature of real edge over a few hundred settled trades. Week 2: same fair-value strategy with more data collapsed. A separate mid-week hypothesis (market-making) also failed cleanly. Fleet ended week 2 at roughly negative thirty-three percent of bankroll. The honest research finding wasn’t on the winning side — it was on the losing side. Adding more parameters to Polybot wouldn’t change that. TradingAgents is asking a separable question.

FIG. 03 — WHAT THE FORK ADDS · THE OPERATIONAL LAYER

Six layers the upstream framework doesn’t ship

Same agent graph, intact. The fork makes it a research instrument rather than a tech demo.

01 · Loop

An autonomous loop

Scheduler · watchlist · auto-trader maps ratings to paper orders · allow-list filtering · per-ticker cooldowns · sector caps · cash checks · position manager evaluates open positions every 60s for TP / SL / max-hold. Append-only audit logs.

02 · Brokers

Multi-broker abstraction

Three modes: local Python broker (yfinance fills, JSON-persisted) · Alpaca paper-trading adapter · “shadow” mode running both in parallel with divergence view. Real Alpaca live endpoints are hard-refused at multiple layers.

03 · Dashboard

A local web dashboard

FastAPI backend · React via CDN, no Node toolchain · SVG equity curve · rolling-peak drawdown · win-rate by rating / ticker / model · exit-reason breakdown · LLM cost vs realised P&L joined by run ID. Runs locally; nothing sent to a cloud service.

04 · Codex

Codex OAuth

Runs the engine on a ChatGPT Pro subscription via the Codex backend. LLM cost floor effectively zero if you already have ChatGPT Pro. Token stored encrypted locally. Falls back to the regular OpenAI API if you’d rather pay per token.

05 · Alerts

Multi-channel alerts

Slack · Discord · SMTP email · configurable filter on rating events and order fills · append-only history kept locally. Webhook URLs masked in API responses so a screenshot can’t accidentally leak credentials.

06 · MCP

MCP plug-ins

Registry for adding Anthropic Model Context Protocol servers (Kensho · Aiera · FactSet · Morningstar · LSEG) as analyst tools. Plug-ins advertise category (fundamentals · news · market data · social) · probe endpoint tests credentials.

Honest-by-design touches: every generated report prepends “Research, not advice” and appends a footer with version, commit, provider, models used, run ID, and cost. Closed trades carry the same metadata. 520+ passing unit tests across engine, services, and HTTP routes. The intent: when the system loses money, the journal makes it impossible to pretend it didn’t.

FIG. 04 — HONEST PRIORS · BEFORE RUNNING THIS IN ANGER

Three priors stated before the data starts arriving

The bias of the project: when the data says no, the dashboard says no, the article says no

It might fail too. LLMs are not oracles, and a sophisticated framework around language-model outputs does not change the underlying error rate of the model. Sample is still everything. The framework’s outputs are subject to the same statistical noise as any prediction system over small samples.

Highest likelihood

If it appears to work, the most likely explanation is variance. The same trap that caught the first article’s candidate edge applies here. A high win rate over fifty trades means much less than it looks. Without out-of-sample confirmation, a flattering early sample tells you almost nothing about whether the system has real edge.

Second-most likely

If it appears to work for the right reasons — empirical win rate matches stated confidence, and alpha-versus-benchmark persists across non-overlapping samples — that would be a meaningful research finding. Whether that happens, I don’t know. The point of putting it in the open is that the data will say.

Genuinely open

This is explicitly not a launch announcement for a product anyone should connect a real brokerage account to. The Alpaca live endpoints are hard-refused at multiple layers in the code, and the design choice is deliberate. The right next step is data, not deployment. The bias of the whole project is straightforward: when the data says no, the dashboard says no, the article says no, and no one tries to retroactively rescue the thesis. That’s the contribution.

FIG. 05 — WEEK THREE · WHAT THE METHODOLOGY WILL MEASURE

Four concrete measurements before publishing findings

The hope: write the week-three article from a position of “here’s what the data says”. The fear: another candidate falsified at higher sample. Both outcomes are publishable.

M1 · Sample discipline

Small watchlist for a few weeks before publishing

A handful of tickers across two or three sectors. Long enough to gather sample, narrow enough to keep attention on what’s actually happening per agent. Avoid the noise of a 65-ticker autonomous loop until the smaller version has been read carefully.

M2 · Calibration view

Stated confidence vs. realised win rate

When the system says “75% confident”, do the trades actually win 75% of the time? Same measurement applied to Polybot’s fair-value model. If the model is systematically over-confident, that bias dominates everything downstream.

M3 · Cost accounting

Cost per ticker · per rating · per profitable trade

With Codex OAuth the marginal LLM cost is effectively zero. With the public OpenAI API, each run is hundreds of agent turns. The honest question: does this scale economically if you ever did run it at real cost?

M4 · Non-overlapping windows

Alpha vs benchmark · out-of-sample

Not within-sample alpha — trivially inflatable. Hold out one period entirely, run the system on the next, then check whether the held-out result matches the in-sample stats. If they diverge sharply, the in-sample was curve-fit.

Open under Apache-2.0 with upstream cited from every relevant surface. Not open: the operator’s running results, the specific watchlist, the per-agent prompt customisations, the alert channels, the trade journals — kept local for the same reason Polybot’s per-experiment data is kept local. Publishing exact configurations encourages people to copy them with real money, which is the opposite of what an honest research project should do. Summary findings will be published. Recipes will not.

The bet is on a different mechanism, not a different parameter setting. The point is not to find a money-printing AI. The point is to put honest measurements of these systems into the public record — so the next person looking at the space starts a step further along than the last.

Thorsten Meyer AI · Introducing Forezai · TradingAgents · § 03

Source dossier & project notes

Forezai · TradingAgents fork — Apache-2.0 license preserved · GitHub published under same author as Polybot · upstream cited from README · marketing pages · generated report footers · forezai.com
TradingAgents upstream — TauricResearch team · multi-agent stock-research framework on LangGraph · the agent graph is intact in the fork · upstream license requirements (notice file · attribution · patent-grant clauses) preserved if anyone forks again
Polybot · Week 1 — 21 parallel strategy experiments on Polymarket 5-minute Up/Down · early winners mostly mechanical illusions · one BTC fair-value taker showed mathematical signature of edge over a few hundred settled trades
Polybot · Week 2 — fair-value strategy collapsed at higher sample · market-making hypothesis failed cleanly · fleet ended at ~negative 33% of bankroll · 78% fleet-wide win rate combined with deeply negative P&L · “you can win four out of five trades and still go broke”
Architecture · Stage 1 — Four analysts in parallel: market structure · news + insider · fundamentals (balance sheet + earnings) · social-media sentiment · each writes a short structured report independently
Architecture · Stage 2 — Bull-bear debate: two researcher agents argue opposing theses from the analyst reports · a research-manager agent arbitrates and writes a single synthesis
Architecture · Stage 3 — Three-voice risk team: aggressive (upside, accepts variance) · conservative (downside, protects capital) · neutral (balances) · forces explicit articulation of downside before any order is proposed
Architecture · Stage 4 — Two-layer decision: trader agent produces three-tier proposal (buy / hold / sell) · portfolio-manager agent synthesises into final five-tier rating with price target and time horizon · PM sees arguments only, never raw data
Operational layer — Autonomous loop · multi-broker abstraction (local Python / Alpaca paper / shadow mode) · FastAPI + React-via-CDN web dashboard · Codex OAuth on ChatGPT Pro · multi-channel alerts (Slack · Discord · SMTP) · MCP plug-in registry (Kensho · Aiera · FactSet · Morningstar · LSEG)
Engineering baseline — 520+ passing unit tests across engine, services, HTTP routes · honest-by-design touches: “Research, not advice” prepended to every report · footer with version, commit, provider, models used, run ID, cost
Hard-refused live trading — Real Alpaca live endpoints hard-refused at multiple layers · operator must deliberately override the refusal in more than one place to actually risk real money
Week three methodology — Small watchlist · calibration view (stated confidence vs realised win rate) · cost accounting per ticker / rating / profitable trade · out-of-sample alpha across non-overlapping windows
Stance — Summary findings published · recipes kept local · the bias is straightforward: when the data says no, the dashboard says no, the article says no, no one tries to retroactively rescue the thesis

Colophon

Set in Source Serif 4 (display, italic accent), EB Garamond (body), IBM Plex Sans (UI labels), IBM Plex Mono (mastheads, ticker, stamps, plug-in tags). Paper-cool gray-cream #e6e7e4.

Chromatic register: structural-slate dominant (committee-architecture analysis), empirical-clay for the Polybot prior forensic, labor-rose for the cautionary “expect to lose money” stance and the regulatory disclaimers, transition-bronze for the week-three forward-shape methodology, alternative-sage for the open-source / honest-measurement positive signal.

Key frame:

FOREZAI / TRADINGAGENTS 13-AGENT COMMITTEE POLYBOT PRIOR OPERATIONAL LAYER EXPECT TO LOSE MONEY CALIBRATION VIEW BULL-BEAR DEBATE HARD-REFUSED LIVE APACHE 2.0 HONEST MEASUREMENTS

Implications for AI-Driven Market Research

This development demonstrates a novel approach to using AI for complex decision-making in financial markets, emphasizing collaborative reasoning among specialized models. It provides a testbed for understanding how AI can simulate or augment trading strategies without risking capital, which could influence future research in AI-based finance and automated decision-making.

While the system does not predict markets or promise profitable trades, it offers insights into the potential and limitations of AI reasoning in structured, multi-agent environments, highlighting the importance of explicit articulation and debate among models.

Amazon

paper trading simulation software

As an affiliate, we earn on qualifying purchases.

Background on AI in Financial Simulations

Previous research, including the Polybot experiments, revealed that many parametric trading strategies fail to survive real-market conditions, often collapsing after promising backtests. This underscored the challenge of developing reliable AI-driven trading systems.

The current project builds on this by shifting focus from rule-based strategies to multi-agent AI systems that reason through structured debate, aiming to see if collective AI judgment can outperform random chance in paper trading scenarios. The concept is part of broader efforts to understand AI’s role in financial decision-making without risking actual capital.

“The TradingAgents framework is a serious step toward understanding how multiple AI models can collaborate to make informed trading decisions in a simulated environment.”
— Thorsten Meyer, researcher at ThorstenMeyerAI.com

Amazon

AI trading decision analysis tools

As an affiliate, we earn on qualifying purchases.

Uncertainties About Real-World Applicability

It remains unclear whether the AI committee’s paper-trading decisions will translate into meaningful insights or profitable strategies outside controlled simulations. The system is designed solely for research, and its effectiveness in live trading or real market prediction has not been demonstrated.

Furthermore, the extent to which multi-LLM collaboration can outperform simple models or human traders under real conditions is still unknown, and the project does not claim to provide financial advice.

Amazon

market research dashboard for traders

As an affiliate, we earn on qualifying purchases.

Next Steps in AI Trading Research

Future developments will likely include extended testing of the TradingAgents system over longer periods and diverse market conditions, as well as refining the agent roles and debate structures. Researchers may also explore integrating additional data sources and enhancing the system’s ability to articulate reasoning.

There is no indication yet of plans to deploy the framework in live trading, but ongoing experiments aim to evaluate its potential as a research tool for understanding AI decision-making dynamics in finance.

Amazon

algorithmic trading paper trading platform

As an affiliate, we earn on qualifying purchases.

Key Questions

Can this system make profitable trades in real markets?

No. The current system is designed for research and runs paper trades; it does not execute real trades or provide financial advice.

How does the AI determine trading decisions?

The system employs multiple specialized LLMs that analyze market data, debate opposing theses, and synthesize their reasoning into trading proposals, all explicitly articulated.

Is this system intended for actual trading use?

No, it is a research platform aimed at understanding AI decision-making in simulated trading environments. It is not configured for live trading with real money.

What are the limitations of this approach?

Its effectiveness in real markets remains unproven, and the system’s decisions are based on simulated data and reasoning structures, not predictive accuracy.

Source: ThorstenMeyerAI.com

Introducing Forezai · TradingAgents — a committee of LLMs decides paper-trades

Up next

Week Three — Foundation model vs Brownian motion. Kronos on five-minute BTC.

Author

Simple Mondays Team

Share article

Introducing Forezai · TradingAgents.
A committee of LLMs
decides paper-trades.