Aurora researches your idea, builds strategies, and backtests the alternatives. You decide what to paper trade and what to take live.
Have an idea? Choose your goal and get a plan with Aurora. Want a starting point? Fork an open-source strategy template and stress-test it yourself. Paper trade first.
Turn the thesis into rules a backtest can check.
Backtest combinations of lookback, cooldown, and allocation.
Rank on Jan 2021–Apr 2025. Test the leaders on 13 months they never saw.
Roll the test window forward before trusting a winner.
Paper trade, monitor, then go live through your broker.
Ask a question in plain English. Aurora runs the research loop and leaves every step open for you to inspect: the rules, each backtest, and the periods every candidate was tested on.
Aurora gathers filings, prices, and fundamentals, then states your idea as rules a backtest can check.
It builds strategy variants and backtests them side by side across lookbacks, timing, and position sizes.
Candidates are ranked on one period, retested on a held-out period, then checked on rolling walk-forward windows.
Paper trade the survivor first. Live trading connects to Alpaca, Public, TradeStation, and Tradier, and you turn it on.
Testing on unseen data makes an overfit strategy easier to catch. No test guarantees future returns, including this one.
Read the thinking. Watch the process. Find an example and make it your own.
Research stories, strategy breakdowns, and lessons from putting investing ideas to the test.
Follow product walkthroughs and trading experiments, from the first question to the results.
Browse worked strategies, inspect their backtests, and try the next step with Aurora or an SDK.
Aurora compared 96 configurations and ranked them on the training period. These six leading candidates then faced an unseen validation period. The encouraging results still need a rolling walk-forward test.
Sortino measures return relative to downside volatility; higher is better. Six displayed candidates from a 25-row leaderboard, ranked on the training period, rather than validation. Returns under each candidate cover the validation window only. Every candidate's validation Sortino is higher than its training Sortino. One later window can do that when the market is easier, so the gap is not proof the strategy improved.
Training: Jan 1, 2021–Apr 19, 2025. Validation: Apr 20, 2025–May 17, 2026. Each candidate varies the lookback, entry cooldown, and position allocation. This is a historical research example, not a live bot record or a forecast.
Candidate 01: 60-day lookback · 60-day cooldown · 15% allocation
Candidate 02: 90-day lookback · 45-day cooldown · 10% allocation
Candidate 03: 90-day lookback · 60-day cooldown · 20% allocation
Candidate 04: 120-day lookback · 30-day cooldown · 10% allocation
Candidate 05: 60-day lookback · 21-day cooldown · 15% allocation
Candidate 06: 90-day lookback · 45-day cooldown · 20% allocation
The founder connected his own brokerage account. You can inspect the portfolio, the decisions, and the mistakes along the way.
Inspect the public portfolioFork a public strategy, inspect the record, and stress-test it yourself. Performance and risk stay on the card.
Browse templatesMax drawdown −15.1% · Sortino 2.89
Max drawdown −12.6% · Sortino 2.48
Max drawdown −32.7% · Sortino 1.71
Records are labeled by deployment type. Updated Oct 11, 2026. Past performance does not guarantee future results.
You did an excellent job of considering all aspects of trading and research which made the application smartly transition from research to strategy, to watchlist, to trade… very nicely done.
It’s not just another AI language model; it’s a visual demonstration of how AI can be practically applied.
It allows me to create my own trading strategy easily and quickly. Even though I know how to code, it’s still very enjoyable to do this with such ease.
Explore and build manually for free. Choose a paid plan for ongoing Aurora research and higher limits. Research tokens are what those runs spend: a short question costs a few, and a deeper model costs more. A daily stock backtest spends no research tokens. A parameter sweep starts at 2 tokens and rises with the number of configurations and the length of the window.
Save ~20% vs monthly*
Learn the platform. Build by hand.
Free to explore. No subscription fee.
Start at your own pace.Start for freeFind and test your next idea.
$480 billed annually
Save ~20% vs monthly · ~34% vs weekly
Get startedMore capacity for serious research.
$960 billed annually
Save ~20% vs monthly · ~38% vs weekly
Get startedThe platform at its highest capacity.
$1,920 billed annually
Save ~20% vs monthly · ~26% vs weekly
Get started*Annual prices are monthly equivalents. The full annual amount is charged once per year. Savings compare the annual charge with 12 monthly payments or 52 weekly payments for the same plan.
The guarantee covers platform subscription fees, not trading losses. Read the policy. Template subscriptions may have separate fees.
No. The three-question start asks about your experience, priorities, and interests, then suggests ideas to test. Choose one or bring your own. Aurora works in plain English, and you can also build rules manually.
Yes, if you choose to. Paper trading needs no brokerage. Live trading connects to Alpaca, Public, TradeStation, or Tradier. Stocks, options, and crypto are supported. The brokerage decides which of those you can trade. Automated live trading requires approval, current consent, and automation you explicitly enable.
You can fork strategy templates and build strategies manually. The free workspace includes 2 paper portfolios with up to 10 strategies each. Aurora research uses tokens: paid plans include a recurring allowance, and you can also buy tokens without changing your plan.
The founder’s Public Portfolio Challenge uses real capital. Marketplace records are labeled as paper or live trading; the Buffett sweep is historical backtest research. These are different kinds of evidence, and none guarantees your future returns.
It’s a monthly equivalent for comparison, not monthly installments. For example, the Investor annual plan shows $40 per month, but charges $480 once per year. Choose Weekly or Monthly for those billing periods instead.
No. Some strategy templates are free to fork; others have their own subscription price. A platform plan covers its research allowance and platform limits. A paid template subscription unlocks that template only. It does not include Aurora access or upgrade your platform plan.
See where the evidence takes you.
