Skip to content

Home

Jump to bottom
Abid Ali Awan edited this page Oct 4, 2026 · 16 revisions

BetterTradingAgents is a crew of AI agents that researches stocks for you. Five analysts dig into technicals, fundamentals, news, Reddit vibes, and a price forecast, a bull and a bear argue about it, and a manager makes the final BUY / HOLD / SELL call with the full reasoning attached. Connect a free Alpaca paper account and the calls turn into simulated orders, with an optional autopilot that runs whole sessions on a schedule.

Warning

Educational project, not investment advice. Everything trades on a paper account; nothing here touches real money.

Start here (about 5 minutes)

  1. Install Python 3.12+ and uv.
  2. Grab the code and dependencies: 
    git clone https://github.com/kingabzpro/BetterTradingAgents.git
cd BetterTradingAgents
uv sync
  3. Get one LLM key (pick whichever looks easiest on Providers). No key yet? Skip this: the app runs in clearly labeled mock mode without one. When you're ready for the full experience, grab the free data keys on the same page.
  4. uv run setup answers a few questions and writes your .env. Then uv run app.
  5. Open http://127.0.0.1:8000, click a quick-add chip like NVDA, AMD, META, hit Analyze Stocks, and watch the agents work.

From there: read the Usage walkthrough, make the calls yours in the per-ticker manager chat, and poke around Configuration when you want to tune things. Everything can also be changed later in the app's Settings page, no file editing and no restarts.

Home screen

Pages

Page What it covers
Providers Where to get every key, with sign-up links and what's optional
Usage Screen-by-screen walkthrough with screenshots: analysis, results, manager chat, incognito mode, history, paper trading, autopilot, notifications, experimental features
Configuration Every environment variable, recommended models, per-role model splits
API REST endpoints, the full result schema, SSE events
Architecture How each subsystem works, module by module
Backtesting Replay the pipeline at past dates and grade it honestly
Testing The offline check catalog
Roadmap What shipped and what's next

Why trust the design?

You shouldn't, until you've seen the receipts. The multi-agent shape improves on the TradingAgents paper, the discovery screen's momentum score follows the published literature, and every call the app makes is graded against what the market actually did 21 days later. The curated reading list lives on the Roadmap.

Home

README · Issues

Clone this wiki locally