README

TradingAgents: Multi-Agents LLM Financial Trading Framework


  
  
  
  




  


 Latest: Trading-R1 is now in early access

After so many requests from the community, Trading-R1 has officially launched!


  


Register for early access ↗

 News


- [2026-11] Trading-R1 v0.1.0 released, and early access is open.
- [2026-10] TradingAgents v0.7.1 released with decisions scored only on final closes and settled across long market holidays, a fix for model dates that could read filings made after the run date, and fixes to portfolios, prices, SEC statements and reports.
- [2026-10] TradingAgents v0.7.0 released with runs on your own ChatGPT Plus or Pro plan, backtests scored from the first close after each decision, a note at start-up when a newer release is out, and a wheel attached to each GitHub release.


Earlier news

- [2026-10] TradingAgents v0.6.0 released with reports saved as one HTML page, a provider per model tier so the managers and analysts can run on different models, past decisions settled for every ticker while the analysts work, and company news read from Yahoo search while Yahoo's news feed is down.
- [2026-09] TradingAgents v0.5.2 released with parallel analysts for a faster analysis, a CLI that runs without prompts from flags such as --ticker and --date, the run's settings recorded in every report, and backtests that see only data published by each analysis date.
- [2026-09] TradingAgents v0.5.1 released with a package layout organised by what each module holds (import paths moved), optional Jev screening of social posts, GPT-6 Sol and Luna as the default models, and fixes to run isolation and SEC EDGAR statements.
- [2026-09] TradingAgents v0.5.0 released with point-in-time integrity across every dated path, SEC EDGAR fundamentals served as filed, backtesting over a ticker and date grid, portfolio-aware runs, and current model lineups across every provider.
- [2026-08] TradingAgents v0.4.0 released with look-ahead / point-in-time fixes across FRED macro, social sentiment, and the decision-log memory; clearer decision signals; working CLI checkpoint resume; Trader price grounding; and the GPT-5.6 and GLM-5.3 models.
- [2026-07] TradingAgents v0.3.1 released with correctness and stability fixes: Alpha Vantage look-ahead filtering, graph-router crash-safety, graph-shape-aware checkpoint resume, working crypto sentiment sources, a configurable LLM retry budget, Bedrock API-key auth, and Claude Sonnet 5 / Fable 5 support.
- [2026-06] TradingAgents v0.3.0 released with a verified data-access contract, an expanded provider registry (NVIDIA, Kimi, Groq, Mistral, Bedrock, and any OpenAI-compatible endpoint), FRED and Polymarket data vendors, a current-generation model catalog, and a CI gate.
- [2026-05] TradingAgents v0.2.5 released with the grounded Sentiment Analyst, GPT-5.5 etc. model coverage, Qwen/GLM/MiniMax dual-region support, TRADINGAGENTS_ env-var configurability with API-key auto-detection, remote Ollama support, non-US alpha

Top issues

Discussions

Latest release

Decisions are scored only on final closes and settle across long market holidays, a model's unpadded date no longer reads filings made after the run date, and fixes to portfolios, prices, SEC statements, reports, Azure and Claude settings.

 Upgrading from 0.7.0

- Docker images no longer include .env.enterprise. For Azure OpenAI in Docker, put its keys in .env.

 Fixed

- A date a model passes without zero padding, such as 2024-7-5, is read as that date. It could compare as later than a filing made after the run date, and SEC EDGAR then served that filing. (1488)
- A decision is no longer sco