Measurements, not marketing claims. Every number on these pages comes from production logs or a reproducible script. Refresh cadence is documented on each sub-page.
Per-endpoint p50, p95, p99 over the trailing 30 days, pulled directly from ApiRequestLogs. No cherry-picking, no synthetic load.
Overhead of calling FlashAlpha through the MCP protocol versus raw REST. 100 identical calls, p50/p95/p99 for both transports.
View MCP-vs-REST benchmark →FlashAlpha's gamma exposure computed from settled OPRA OI compared to a hand-recomputed NBBO reference for SPY. Methodology, sample, deviations.
Planned - not yet publishedThe flashalpha-py / flashalpha-js / FlashAlpha .NET clients each add a thin layer over HTTP. This page will quantify the cost.
Planned - not yet published