TradingAgents v0.7 runs on your ChatGPT plan, hits 115,000 stars

Running an AI trading research framework used to need a developer, a cloud account and an API bill. That stopped a lot of non-engineer traders from trying one. TradingAgents v0.7 removes the API bill. You sign in with a ChatGPT Plus or Pro plan instead.

Tauric Research released v0.7.0 on 16 October. The repository sits at 115,000 GitHub stars this morning. It ranked 17 on Trendshift's daily list on 26 October.

What the project is

TradingAgents is a research framework. It uses several AI agents, each with a job. One agent studies the company's fundamentals. One reads social sentiment. One looks at technical charts. A trader agent takes a decision. A risk management team checks it. These agents share work on a stock before suggesting a trade. The project is open source and has an arxiv paper from December 2024. The repository's disclaimer labels the output as research only. It asks the reader to treat the output as a study signal, nothing more.

What v0.7 changes

The new chatgpt provider in v0.7 runs the whole pipeline on a ChatGPT Plus or Pro plan. The user signs in once from the command line. The sign-in uses OpenAI's authentication for open-source apps. The CLI then lists the plan's models and treats reasoning effort the same way an API account would. The release also attaches a Python wheel to each GitHub release. The install is one step from the release page, with no need to clone the repository.

What the author also fixed

Version 0.7 scores every backtest decision from the first close after the decision date. Older runs scored the decision at the close of the date itself. The project says this stops a run from being scored on news a trader could not have seen. Decisions already settled in a memory log keep the old entry. Mixing a resumed run with a fresh one is not a clean compare. Starting a new run is the clean path. A run without a terminal, like a cron job or CI, no longer calls the announcements server at start-up.

Where the limits are

The subscription provider only shows models that OpenAI exposes through its open-source sign-in. A model only available through the API still needs the API path. The release notes say reasoning effort behaves the same way as on an API account. Model prices and plan quotas sit with OpenAI. Those change with the plan. The README's disclaimer asks a reader to treat the output as a research signal, for study only. The sign-in change matters for one group of traders in particular. They have watched this project pass 100,000 stars. They never had a developer to connect it to an API. Now they can try a run tonight, on the plan they already pay for.

Source

Primary source: TradingAgents CHANGELOG, v0.7.0. Repository: TauricResearch/TradingAgents.