Real-economy and cross-asset feeds, built as clean causal inputs. The signals the crowd isn't already trading.
Price, volume, the standard technical library. When every desk searches the same input space, a beautiful backtest is a statistical certainty, not a discovery. The durable leverage is upstream, in inputs that are causally ahead of price and mostly absent from retail pipelines. That is the half of the problem we build.
PCA and a Marchenko-Pastur cut pull the handful of real forces out of the S&P in a few lines of numpy, and every desk runs the same decomposition. The edge was never in the factors. It is in the residual, the idiosyncratic leftover once the market and sector tides are hedged out, and that residual moves on things a covariance matrix cannot see: an earnings surprise, an insider or congressional buy, an analyst revision, a sanctions listing, a supply or energy shock in one sector. That is the input space we build. Point a feed at your residuals and see if it moves your leftover.
Not sixty-six scattered endpoints. A coherent set of the inputs a systematic process can actually use, grouped by what they measure.
{
"market": "COPPER-#1 - COMEX",
"report_date": "2026-08-04",
"cot_signal": "Crowded Long (contrarian bearish)",
"large_specs": {
"net": 77123,
"net_pct_oi": 26.6,
"net_percentile_3y": 99.1,
"net_wow_change": +9842
},
"commercials": { "net": -86169 },
"open_interest": 289395
}
Large speculators are net long copper at the 99th percentile of their three-year range. Positioning stretched that far tends to mean-revert, and the producers who hedge it sit on the other side. That is a real-economy input, and it is not in your price series.
{
"ticker": "AMD",
"smart_money_signal": "MIXED",
"conviction": "2 of 4 cohorts bullish, 2 bearish",
"insiders": { "signal": "bearish", "detail": "0 open-market buys, 25 sells worth $33.7M (90d)" },
"congress": { "signal": "bearish", "detail": "3 disclosed trades: 1 buy, 2 sells" },
"institutions": { "signal": "bullish", "detail": "6 of top 10 holders added, 0 trimmed" },
"analysts": { "signal": "bullish", "detail": "Buy, 71.4% buy ratings, +29.7% upside" }
}
One call folds four smart-money cohorts into one read. Here, insiders and Congress are selling into a name that institutions and analysts still like, so the composite says MIXED rather than forcing a call. The /demo endpoint runs free and rotates its ticker daily. For any ticker, call /smartmoney?ticker=SYMBOL with a trial key or x402, and you get the same JSON.
No wallet, no sign-up, no pitch. You evaluate real data the way you evaluate everything: does it survive out of sample.
We don't sell you a strategy. We give you inputs the crowd isn't already trading, and you decide if they earn their place.