Why Finny?

Finny started with a simple frustration: the distance between having a trading idea and actually knowing whether it works is far too large. We wanted to close it. This is the short version of what Finny is and why it exists.

The gap we kept hitting

Anyone can describe a trading idea in a sentence — “buy when the 50-day crosses the 200-day,” or “fade the big overnight gaps.” Turning that sentence into something you can actually test is where most people stop. The tools tend to fall into two camps: no-code black boxes where you can't see or trust the logic, and full quant frameworks with a steep on-ramp of boilerplate, data wiring, and backtest plumbing. There wasn't much in between.

What Finny is

Finny is an AI trading assistant that lives in your terminal. You describe an idea in plain English; Finny writes a complete Python strategy, backtests it against real historical data, and helps you refine it — turn by turn, until it either holds up or you've learned why it doesn't.

• The code is yours. Finny generates open, readable Python you can inspect, edit, and keep — not a setting buried in a dashboard.
• Backtests run on real data. The engine_v2 backtester reports metrics across return, risk, drawdown, and stability — not a single cherry-picked number.
• Broker connections are optional and local. Link Alpaca, Binance, or IBKR when you're ready. Credentials stay on your machine.
• Bring your own model. Point Finny at Claude, OpenAI, Gemini, or a local model with your own key.

Why the terminal, why open code

A strategy you can't read is a strategy you can't trust. We made a deliberate choice to keep everything Finny produces transparent and editable: the generated Python, the backtest results, and the broker configuration all sit in plain files on your machine. The terminal keeps you close to that work instead of abstracting it away behind a UI you don't control.

Who it's for

Finny is built for traders who can describe an edge but don't want to hand-write the scaffolding around it, and for developers who want an AI collaborator that respects the command line. If you'd rather read the code than take a black box's word for it, Finny is for you.

What's next

This is the first post of more to come. We'll write about how the backtester actually works, how research subagents pull the market apart before a single line of strategy is written, and what we're learning building an AI that trades. In the meantime, the docs are the fastest way to see Finny in action.

— Jaimin & Nisarg