I got frustrated with the loop of: have a strategy idea → spend hours writing boilerplate → realize the logic is broken → start over.
So I built Finny — a terminal agent (Claude Code, for financial markets). You describe what you want, it generates a validated Python strategy, and you backtest it immediately against historical data.
I've built 3-4 algorithms with it myself. They're currently live, and I'm sitting at +2.3% net this week.
Three agent modes: Build (generates immediately), Research (analyzes first), Chat (Q&A). Tab to switch mid-session. Supports any asset yfinance covers — crypto, stocks, ETFs. BYOK — Anthropic, OpenAI, Google, or local models.
npm i -g finny && finny
Happy to answer questions about the architecture or validation layer. Still early, curious what strategies people would throw at it.
1 comment
[ 9.0 ms ] story [ 879 ms ] threadSo I built Finny — a terminal agent (Claude Code, for financial markets). You describe what you want, it generates a validated Python strategy, and you backtest it immediately against historical data.
I've built 3-4 algorithms with it myself. They're currently live, and I'm sitting at +2.3% net this week.
Three agent modes: Build (generates immediately), Research (analyzes first), Chat (Q&A). Tab to switch mid-session. Supports any asset yfinance covers — crypto, stocks, ETFs. BYOK — Anthropic, OpenAI, Google, or local models.
npm i -g finny && finny
Happy to answer questions about the architecture or validation layer. Still early, curious what strategies people would throw at it.
GitHub: https://github.com/Jaiminp007/finny