Insider Buy Signals

We read Form 4 filings from SEC EDGAR and keep only two well-defined kinds of open-market buy — cluster buys (two or more insiders in the same company within seven days) and buys that are large relative to the insider's existing holdings. Option exercises, 10b5-1 plan sales, and sub-threshold noise are dropped. This is a clean, filing-verified filter over public purchases, not a prediction: we tested whether these two filters precede price moves over our own history and could not find a return edge distinguishable from base rate, so we don't claim one. What you get is the real money-on-the-line buys, fast and linked to source — you decide what they mean.

Every field we return — shares, price, dollar value, percent increase versus holdings — is parsed from a real filing and ships with a link to that filing on sec.gov. Records that can't be linked are dropped, not guessed. No scores, no return forecasts, no unsourced numbers.

A cluster is two or more insiders buying the same company inside a single window of seven calendar days or fewer, grouped across the whole corpus rather than day by day. An earlier build measured that window as seven days on either side of each filing, which could stretch a group to fourteen days end-to-end and overstate how many insiders truly bought inside one seven-day span. That has been corrected to a true ≤7-day window, so a handful of clusters now report fewer insiders and a smaller combined dollar figure than before — the honest count for the window as stated. Which filings cluster is unchanged; only the over-reached insider counts came down.

One line in the Claude Code CLI connects the live MCP server:

claude mcp add --transport http insider-signals https://mcp.tradestarinsider.com/mcp

Hard limits, no overage — requests over the cap return 429 until the window resets. The MCP server is free and rate-limited. Metered API billing runs through RapidAPI.