Open-source · Edge-native

A swing-trading agent you can backtest, paper-trade, and share.

Three daily cycles — EOD, Morning, Intraday — with Kelly-sized entries, deterministic or Workers-AI portfolio manager, and fully-resumable backtests. Publish runs to a public leaderboard so others can see how your strategy holds up.

Cycles per trading day
3

EOD · Morning · Intraday

Backtest engine
Workflows

Resumable, durable

Position sizing
½-Kelly

Static or rolling

PM modes
AI · Deterministic

Workers AI / heuristic

Public leaderboard

Latest top performers

View full leaderboard →

Loading…

What it does

A complete swing-trading loop, end to end

Backtest a strategy, run it against your Alpaca paper account on schedule, then publish the results so they’re part of the public record.

Resumable backtests

Per-day Workflows with object-store checkpoints — deploy mid-run and resume from the last completed day. Compare up to 3 runs side-by-side.

Live paper trading

Alpaca paper account driven by cron. EOD signals at close, morning entries with bracket orders, intraday sells with trailing stops.

Kelly position sizing

Half-Kelly with static (signal) or rolling (in-run / live trade history) win-rate. Tracks Actual % vs Kelly target on every entry.

Two PM modes

Workers AI (Llama 3.1 8B) for explainable EOD selection, or deterministic heuristic mode for fully reproducible runs.

Sharpe, Sortino, drawdown

Excess-vs-SPY Sharpe & Sortino, max drawdown, and trade analytics — computed at session and backtest level, with metric tooltips.

Public leaderboard

Publish backtest, paper, or session snapshots to a no-auth public leaderboard. Strip private data, keep ownership via one-time delete token.

Architecture

Edge-native architecture

Every component runs on the edge. No long-running servers, no third-party schedulers, no manual workflow orchestrator.

Edge runtime
Workers + TanStack Start

Single deploy, SSR, RPC to Durable Objects

State
SQL + Object storage

Sessions, cycles, trades, daily stats · backtest datasets

Backtests
Workflows

Durable per-day steps with checkpoint resume

Portfolio manager
Workers AI · Llama 3.1 8B

or deterministic heuristic for reproducible runs

Auth
Google

Multi-user with per-user Alpaca credentials

Open source
MIT

Self-host on your own infrastructure

Run your first backtest in minutes

Sign in with Google, add your Alpaca paper key, and queue a backtest on a date range. Publish the result when you’re happy with it.