GitHub: https://github.com/Kirby54/Snap-Trade
Release: https://github.com/Kirby54/Snap-Trade/releases/tag/v2.1.0
Why I built it
Most "AI trading" tools give you a signal and a confident tone.
I wanted to explore something different: what if the AI had to justify every trade, let a script check its numbers, and say "no trade" when the chart doesn't show a clear edge?
In Snap & Trade, you paste a screenshot from any platform (TradingView, MT4/MT5, FX Replay, Binance...), for any market and any timeframe. Claude reads the chart step by step, scores the setup out of 5, and gives you LONG, SHORT or NO TRADE with entry, stop loss, TP1, TP2 and invalidation.
What's included
- Screenshot analysis for stocks, crypto, forex, indices and futures
- A confluence score (trend, level, candle, momentum, volume) instead of fake win rates
- A Python script that checks every plan and refuses anything under 1.5R
- Position sizing from your account and risk %
- A backtest journal: win rate, average R, profit factor, losing streaks
- 11 modules from day-trader-skill by Jake Schincariol (MIT): scanner, news, SEC earnings, risk gate, Alpaca orders (paper by default), alerts, and more
- 324 tests, with CI on Linux, Windows and macOS
The stack is plain Python (standard library only) and Claude skills.
One design decision I'm especially interested in
The LLM never does the math. In my first test on a Nasdaq-100 chart, the obvious breakdown short only paid 0.91R. The script refused it and suggested a better limit entry on the retest instead.
What's not included yet
This is not financial advice, and there is no proven track record yet. I'm running a bias-free backtest on FX Replay right now, and the results will go in the README, losses included. The live-data modules only cover US stocks, and the skill never places an order without an explicit yes from you.
Try it
/plugin marketplace add Kirby54/Snap-Trade /plugin install snap-and-trade@snap-and-trade Then paste a chart and ask "long or short?".
GitHub: https://github.com/Kirby54/Snap-Trade
I'd love technical feedback: would you trust an AI trade plan more if it was forced to show its reasoning and could refuse to trade?

