Deutsch | Español | français | 日本語 | 한국어 | Português | Русский | 中文

Trading-R1: The terminal for financial research and analysis

Quick start

1. Create a Tauric account and get an API key. Sign up, then create a key.

2. Install.

curl -fsSL https://raw.githubusercontent.com/TauricResearch/Trading-R1/main/install.sh | sh

Or, with Python 3.12 or later and bun: pip install trading-r1.

3. Add your key and start.

export TAURIC_API_KEY=your-key
trading-r1

To store the key instead, run trading-r1 secrets set TAURIC_API_KEY, or press / in the app and choose API keys….

To uninstall, run trading-r1 daemon uninstall if you installed the service, then uv tool uninstall trading-r1 or pip uninstall trading-r1. Your data stays in ~/.trading-r1 and your reports folder.

Choose a model

Name	Model	Needs
tauric	tauric	TAURIC_API_KEY
gpt	gpt-6.1-sol	OPENAI_API_KEY
anthropic	claude-sonnet-5-5	ANTHROPIC_API_KEY
gemini	gemini-3.8-flash	GEMINI_API_KEY
vllm	Qwen3.8-27B	a vLLM server you run

The app starts on the first you have set up, in the order OpenAI, Anthropic, Gemini, Tauric, vLLM. Change it with Model… in the palette, or start on one with trading-r1 --profile gpt. To point vllm at your server, see Configuration.

Use it

trading-r1                                          # open the terminal
trading-r1 "What's NVDA's last quarterly revenue?"  # and ask a first question
trading-r1 --mode analytical                        # start in a mode

Key	Does
/ or ^K	Open the command palette
shift+tab	Cycle the mode
^N	Start new research
esc	Cancel the running turn, or close a dialog
^C	Quit

Mode	Focus
auto	Full toolkit; the agent decides
research	Web first, for deep research
analytical	Numbers first: data and valuation tools, with the web for context

The palette lists every action: sessions, model and mode, reports, channels, scheduled reports, tools and skills, API keys, the background service, Settings and Help.

Architecture

Reports and sending

Export report… saves the last answer as a PDF and, by default, Markdown, in ~/Documents/Trading-R1 or ~/Trading-R1.

Send report… sends it to your channels, and sets one up if there is none.

Channel	You give	Gets
telegram	a bot token from @BotFather, then write to the bot once and confirm the chat	text and the PDF
email	your address and an app password (presets: gmail, icloud, yahoo, fastmail)	text and the PDF
folder	a path	the report files

To ask from your phone, set the Telegram channel's reply to on. While the backend is up, your bot answers your messages with a short research run, and /new starts over.

Scheduled reports

Jobs ask a question on a schedule and send the report to your channels. Use Scheduled reports… in the palette, or:

trading-r1 jobs add --name "Morning brief" --every weekdays --time 08:00 --tz America/New_York \
    --prompt "What moved my watchlist overnight? NVDA, AAPL, MSFT." --deliver email,telegram

The commands and options for reports, channels and jobs are in the manual.

Background service

Jobs and Telegram replies need a running backend, and the terminal's stops when you close it. To keep one running, install a user service (Linux with systemd, macOS with launchd), from Background service… in the palette or:

trading-r1 daemon install      # also: status, logs, start, stop, restart, uninstall

After you update, or change a key in your shell profile, run trading-r1 daemon install again.

Manual

Commands, settings, session sync, your data and keys, configuration and adding a skill are in the user manual. To work on the code, see CONTRIBUTING.

Citation

@article{xiao2025trading,
  title={Trading-r1: Financial trading with llm reasoning via reinforcement learning},
  author={Xiao, Yijia and Sun, Edward and Chen, Tong and Wu, Fang and Luo, Di and Wang, Wei},
  journal={arXiv preprint arXiv:2509.11420},
  year={2025}
}

Disclaimer

For research and education only, not investment advice.

License

Apache-2.0. See LICENSE.