Dev.to · 5 min read

How to track congressional stock trades with Claude, and what the disclosures leave out

How to track congressional stock trades with Claude, and what the disclosures leave out

All outputs are real tool calls to the Equibles MCP server from 21 October 2026, trimmed. Every congressional trading tracker shows the same table: who bought what. The table is the easy part. The hard part is reading it: when the trade happened versus when it was disclosed, how much money a "$1,001 to $15,000" row really is, and whether a row is stock at all. Here is how I ask an AI agent those questions, and the checks I run on its answers. Connect it first In Claude, open Customize → Connectors → Discover, search for Equibles, click Connect to Claude and Authorize. Then switch it on in a chat from + → Connectors. In ChatGPT it's Plugins, search Equibles, Install plugin. In Claude Code: claude mcp add --transport http equibles https://mcp.equibles.com/mcp then /mcp to sign in. The free plan covers 100 requests a day. Other clients are in the setup docs. 1. "What has Congress been buying?" The prompt: "Using Equibles, which stocks did members of Congress buy most in trades disclosed over the last 90 days?" The agent calls GetMarketWideCongressionalActivity with days=90: # Ticker Buyers Trades (buy/sell) Est. net flow Top buyer Latest trade 1 BE 1 4/0 +$7.5M Nancy Pelosi 2026-07-28 2 INTC 2 5/4 +$1.02M Nancy Pelosi, Thomas H Tuberville 2026-07-24 3 MRK 1 2/1 +$57K Maria Elvira Salazar 2026-09-24 4 BRK-B 2 4/0 +$56.5K Maria Elvira Salazar, Richard Dean McCormick 2026-08-07 5 BWXT 1 4/0 +$56.5K April McClain Delaney 2026-08-03 Three things to know before reading anything into it: The flow is an estimate. Members disclose a range, not an amount. The tool counts each row at the midpoint of its range, so "$7.5M" is a sum of midpoints. The window is the disclosure date. "The last 90 days" means trades filed in that window. The trades can be older; the STOCK Act's general deadline is 45 days after the trade. One member can be the whole row. Bloom Energy's +$7.5M has one buyer. Drill down before calling it "Congress". So the next prompt is "Show me Nancy Pelosi's trades since June" (GetMemberTrades): Trade Filed Ticker Range Asset type Owner 2026-07-24 2026-08-21 BE $1,000,001 to $5,000,000 ST Spouse 2026-07-24 2026-08-21 BE $1,000,001 to $5,000,000 OP Spouse 2026-07-24 2026-08-21 INTC $500,001 to $1,000,000 ST Spouse 2026-07-24 2026-08-21 INTC $250,001 to $500,000 OP Spouse 2026-07-28 2026-08-21 BE $500,001 to $1,000,000 ST Spouse 2026-07-28 2026-08-21 BE $500,001 to $1,000,000 OP Spouse The $7.5M is four rows. Two of them are options (asset type OP, not ST for stock), the owner column says Spouse on all of them, and they were filed 24 and 28 days after the trades. A headline like "Pelosi bought $7.5M of Bloom Energy" makes three claims the filing doesn't: about who, about what, and about how much. 2. "Who in Congress trades this stock?" For one ticker, GetCongressionalTrades lists every disclosed trade. Snowflake (SNOW) over the last year: Trade Filed Member Type Range Account 2026-09-24 2026-10-07 Maria Elvira Salazar Sale $1,001 to $15,000 UBS IRA Account 2026-09-04 2026-10-07 Maria Elvira Salazar Purchase $15,001 to $50,000 UBS IRA Account 2026-09-04 2026-10-07 Maria Elvira Salazar Purchase $1,001 to $15,000 UBS Brokerage Account 2026-08-07 2026-09-10 Gilbert Cisneros Purchase $1,001 to $15,000 150 Main Street Trust > Bank of America The purchases were filed 33 and 34 days after they happened. That lag is the whole story for anyone who wants to copy these trades: by the time the row exists, the price it was made at is a month old. I ask the agent to add a "days to file" column to every answer like this one, because it changes how the table reads. The same data has a page per stock, for example SNOW's congressional trades. 3. "What does this member trade?" The 30-day version of the leaderboard put ON Semiconductor first, with an estimated $32.5K. That was a single row: Gilbert Cisneros' $15,001 to $50,000 purchase on 28 September. Asking for all his trades since 1 September returns 51 rows: all 51 filed on the same day, 12 October; 27 purchases and 24 sales, across 46 tickers; 43 of them in the lowest band, $1,001 to $15,000; every one in the same trust account at Bank of America. A leaderboard ranks rows. Whether a row is a considered bet or one line of a 51-line batch is something only the drilldown shows, and the disclosure doesn't say who chooses the trades. That's why I never let an agent stop at the leaderboard. The prompt I reuse Using Equibles: 1. List the stocks Congress bought most in disclosures from the last 90 days. 2. For the top 3, show every underlying trade with trade date, filing date, amount range, asset type, owner and account. 3. Add the days between trade and filing, flag options and spouse accounts, and say how many members are behind each row. What the disclosures can't tell you Exact amounts. Only ranges, so every dollar total is an estimate. Why. A filing has no rationale, and a trust or managed account may not reflect the member's own choices. Timing. A month is normal, and late filings arrive long after the trade. Results. Whether a trade worked isn't in the filing; that needs prices, which the agent can fetch, but it's a separate question. The market-wide view is on the Congress leaderboard if you'd rather browse than prompt. Nothing here is investment advice.

This is a summary aggregated from Dev.to. Read the complete article on the original site:

Read full article at Dev.to

More AI & Machine Learning News