AFFLATUS
afflatusNounA Divine Breath Of Inspiration
The Spark Behind Every Build
Describe Your Trading Idea In Plain Words. Afflatus Specs It, Codes It And Walk-Forward Tests It On Up To 25 Years Of Historical Data.
TradingView Pine·MetaTrader 5·NinjaTrader
Data Varies By Market And Timeframe. Hypothetical Backtests. Not Investment Advice. Past Performance Is Not Indicative Of Future Results.
Ephesians 1:17–18
Stop Guessing. Start Testing.
Brain-dump a trading idea in your own words. Afflatus agents turn it into a spec, write the code, backtest it and put it through the tests most ideas fail. You get a plain verdict, with the reasons.
Every trader has more ideas than time to test them. Afflatus takes the idea you would have shelved and gives it a real test in minutes, with the same rules for every idea, so you can keep the good ones and let the rest go without regret.
Where do your trading ideas live right now? A screenshot, a hunch, a note on your phone? Scattered, untested, half-remembered.
The ones that do get tested are often tested badly. Signals that peek at the next bar, fills that ignore the spread, the best result picked from a hundred tries.
So traders trust curves that were never real, and drop ideas that never got a fair look.
Afflatus gives every idea the same careful process, from the first sentence to the final verdict.
01 Brain-Dump. Describe the idea the way you would tell a friend. Half-formed is fine. That is what the next step is for.
02 Spec. An agent asks the questions a careful tester would: which market, which timeframe, what exactly triggers an entry, where you are wrong, when you leave. Your idea becomes a precise spec.
03 Code. Four languages in one: the logic is built in Python first, then translated into Pine Script, MQL5 and NinjaScript for NinjaTrader 8. You get readable code you can open on your own platform.
04 Backtest. The spec runs on our engine over the Quant Data Farm: up to 25 years of market history from established market data vendors (varies by market and timeframe), with spread, commission and slippage charged on every trade. Signals act on closed bars only. Bring your own market data and run the same test on yours and ours, to verify our data against data you already trust. Share it after review and the whole community grows the farm together.
05 Verdict. A plain answer: Passed Historical Tests, Did Not Pass, or Inconclusive. Always with the reasons, and suggestions for further testing.
Twin Check. We run your logic on a random-walk twin of the market. A fair test should find nothing there. If it does, something is peeking at the future.
Zero-Edge Null. Same number of trades, same holding times, same stops and targets, but random entries. Your idea has to beat chance by a clear margin, not just look profitable in a backtest.
Flip Test. Every buy becomes a sell. If the flipped version does as well, the logic spots that a move is coming, not which way. That is a volatility filter, not an edge.
Out-Of-Sample. Settings are chosen on one stretch of history and judged on a stretch they never saw. Whatever looked best in the past tends to fade, so the past is not where we judge.
Monte Carlo. We resample your trades 5,000 times to see the paths the same logic could have taken: the unlucky streaks, the deepest drawdowns, the worst runs in a row. One lucky order of trades cannot carry a verdict.
Kairos Trust Score. Our own timing test, scored on out-of-sample trades only: does an entry move your way before it hurts you? It weighs stop safety, how much of the move is captured, drawdown and win quality into one score out of 100. It asks whether entries come at the right moment, not just eventually.
Calmar Ratio Test. Return per unit of worst drawdown, measured three ways: on the raw trades, across all 5,000 Monte Carlo paths, and on the $500 seed. Below 1, the idea earned less than its own worst fall.
Daily Seed Test. Every account-level test starts from a $500 seed. No idea should need more than $500 at risk to show what it can do, so $500 is the baseline we measure against. Results read as multiples of the seed, not dollar amounts, so the same test reads the same way on a larger account.
Floor Test. Thousands more resampled paths on the $500 seed with a hard $255 floor. Every path that breaks the floor is counted, and that breach rate weighs heavily in the Stackable Factor rating: the more often an idea breaks the floor, the lower it rates.
The Full Quant Battery. Every run also reports a t-test and bootstrap confidence interval on the average trade, the Probabilistic Sharpe Ratio corrected for skew and fat tails, the minimum track record needed for confidence, Sortino, SQN, Ulcer index, tail ratio, skew and kurtosis, and each trade’s worst and best excursion. No verdict rests on one headline number.
Passed Historical Tests means the idea survived our tests on simulated past data. Results are hypothetical, are not a recommendation, and do not mean the idea will make money.
Trained On How We Work.
Each agent does one job in the pipeline, from the first question to the final verdict. They carry the methods Kingdom Portfolios uses on its own research, including the tests that caught our own mistakes.
Our HYPO-TEST-IFYING Flow
Every great algo starts as a spark. Shape a team of agents that grows your idea step by step into something clear, tested and ready to build on, then take the whole team home as one file for your own Claude.
HYPO Hypothesize
Speak the idea the way it lives in your head. Your Discovery and Edge agents listen closely, ask the curious questions, and help you name exactly what it looks for and when it acts. Each answer sharpens the picture until your hunch becomes a precise, testable plan.
TEST
Your Risk, Build and Validation agents turn the plan into real code and run it across deep market history, with spreads and costs on every trade. Out-of-sample data, a twin market and a random baseline show you clearly how your idea would have behaved, so every result you keep is one you understand.
IFY Justify
Your Optimization, Cabinet and Records agents refine your favourite version, find broad plateaus of settings, and show how your idea pairs with others in a cabinet. Every version is filed with its reasons, so your library grows wiser with each idea you bring.
Each pass teaches your agents more about how you think. The more you bring, the more your team understands your style.
Training means instructions, notes and worked examples your own Claude follows, not model fine-tuning. Afflatus is a research and backtesting tool: results are simulated on past data and are not investment advice. Past or simulated performance does not indicate future results.
Two ways in. Bring the Claude you already use, free to connect, or let Afflatus run everything for you in the browser.
Free · MCP
Add Afflatus to Claude Code or as a custom connector in the Claude app, and brain-dump right where you already think. Your own Claude plan does the writing. Afflatus does the backtests, the tests and the saving.
claude mcp add --transport http \ afflatus https://afflatus.club/mcp
Seamless · Weekly Plan
No setup and no Claude subscription needed. Afflatus brings the AI to you, with a fresh allowance every week, and opens up the fastest data and the full stacked-testing lab.
Start Seamless from Subscription in the app after you sign in. Cancel any time.
The data hub is where every backtest starts: decades of bars across the markets traders actually test, labelled exactly as what they are. During the beta the hub is filling in market by market, and you can always test on your own data.
Everything starts private. Share an algo in the public library with your name on it, or keep it to yourself. Kingdom Portfolios can access what is stored on Afflatus to run, secure and support the service. We will not use, share or sell your algos, or use them in our own trading, without your separate written consent. Every staff view of a private algo is logged.
Run as many of your Afflatus algos as you like. Each one speaks in pure signals: market, direction, risk and targets, never lots or a broker.
The Parallelizer maps each signal to forex and futures using your own market map and sizing rules, sized to each account’s balance.
A floor that rises with the account’s high-water mark and a goal for the day, checked before every order. Your rules. ARM enforces them.
Picture every algo you have tested, working together from one calm cockpit. ARM is designed to keep your rules, your risk and your rhythm moving as one across every account you connect. You stay in command while ARM takes care of the busywork.
ARM is a separate tool that works hand in hand with Afflatus: software for accounts you own, now in its final quarter of beta testing and bug patching, scheduled to open February 11, 2027. It sends no signals of its own and makes no trade calls. Trading forex, CFDs and futures involves substantial risk of loss, and losses can exceed a floor in fast markets.