███████╗██╗███╗   ██╗███╗   ██╗██╗   ██╗
                                    ██╔════╝██║████╗  ██║████╗  ██║╚██╗ ██╔╝
                                    █████╗  ██║██╔██╗ ██║██╔██╗ ██║ ╚████╔╝
                                    ██╔══╝  ██║██║╚██╗██║██║╚██╗██║  ╚██╔╝
                                    ██║     ██║██║ ╚████║██║ ╚████║   ██║
                                    ╚═╝     ╚═╝╚═╝  ╚═══╝╚═╝  ╚═══╝   ╚═╝

Build, backtest, and iterate on trading algorithms using natural language. Built on the OpenCode AI agent harness.

---

Quick Start

curl -fsSL https://finnyai.tech/cli/install | bash
finny

Or via npm (same public package the curl installer uses):

npm i -g finny
finny

Try it: "Build me a momentum-based ETH trading strategy with RSI signals"

---

Agent Modes

Press Tab to switch between modes.

Mode	Description
Build	Generates algorithms immediately. No questions asked.
Research	Asks clarifying questions, fetches market data, runs analysis, then transitions to Build.
Chat	Conversational assistant for strategy questions, portfolio review, market data.

---

How It Works

1. Describe what you want in plain English
2. The agent generates a Python Strategy class (7 built-in templates available)
3. Code is auto-validated — syntax, safety, and trading best practices
4. Algorithm saved to cloud storage (Convex) with version tracking
5. Run backtests against historical data
6. Iterate — modify, re-validate, compare versions

---

Strategy Templates

Template	Description
Momentum	RSI momentum — buys oversold, sells overbought
Mean Reversion	Bollinger Bands — buys at lower band, sells at upper
Breakout	Donchian channel — buys new highs, sells new lows
DCA	Dollar-cost averaging — systematic buying with profit-target exit
Golden Cross	SMA 50/200 crossover — buys golden cross, sells death cross
Scalping	EMA scalping with tight stops — quick entries and exits
Custom	Minimal skeleton — implement your own logic

---

Commands

Command	Description
/algos	List saved algorithms
/code	View algorithm source code
/backtest	Run historical backtest
/review	Review strategy changes (commit, branch, or PR)
/init	Initialize project rules for strategy generation

---

Backtesting

• Historical data via yfinance
• Configurable duration, interval, and starting capital
• Metrics: Total Return, Sharpe Ratio, Max Drawdown, Win Rate, Profit Factor
• Interactive TUI for parameter selection and results display

---

Validation

Every generated strategy is checked for:

• Python syntax errors
• Strategy class structure — strict v2 requires Strategy(broker, params=None) and on_bar(self, symbol, bar)
• Forbidden imports — os, subprocess, socket, requests, etc.
• Dangerous calls — exec, eval, compile, __import__
• Trading pitfalls — lookahead bias, unbounded lists, division by zero

---

Configuration

BYOK — bring your own AI API keys. Works with Anthropic, OpenAI, Google, or local models.

Supported assets include crypto (BTC, ETH, SOL) and stocks (AAPL, NVDA, PLTR).

Config is stored in XDG-compliant directories:

~/.config/finny/    # configuration
~/.local/share/finny/  # data

---

Tech Stack

TypeScript / Bun / Turbo monorepo. Solid.js TUI. Convex DB. Python backtesting engine.

Fork of OpenCode by Anomaly.

---

License

Finny is available for noncommercial use under the Finny Source-Available License 1.0 (PolyForm Noncommercial 1.0.0 with Additional Conditions).

In short: you may use official Finny releases and read the source freely. If you fork or modify Finny, you must publish your changed source publicly within 30 days — private forks are not licensed. Nobody may host Finny, or any fork of it, as a web or network service for others. Commercial or enterprise use requires prior written approval from Finny and a separate commercial license agreement.

The original OpenCode codebase is licensed under the MIT License. Finny's modifications and new features are licensed under the Finny Source-Available License 1.0.

---

Acknowledgments

Built on OpenCode by Anomaly.