About the Sentinel Knowledge Graph
The Sentinel knowledge graph is the largest open dataset of congressional influence ever assembled. As of 2026, the graph contains 33M+ nodes and 7,338,730 edges spanning 67 entity types and 104 relationship types, constructed from 15 federal databases.
The graph integrates data from House Periodic Transaction Reports (PTR), Senate Electronic Financial Disclosures (eFD), Federal Election Commission bulk campaign finance records, congressional committee assignments and legislative histories, Lobbying Disclosure Act filings, Foreign Agent Registration Act (FARA) disclosures, federal court documents via PACER, Department of Justice enforcement records, IRS Form 990 nonprofit filings, and USAID foreign assistance disbursements.
What you can explore
Congressional stock trades: 16,238 TRADED_STOCK relationships link legislators to companies they traded during their tenure. Sentinel has identified 810 closed influence loops — cases where a company donated to a legislator who then traded that company's stock. The most significant loop involves Pfizer and Senator Perdue, with a loop strength score of 100.
Campaign finance: 719,033 DONATED_TO edges trace corporate contributions directly to legislators, revealing which industries fund which members of Congress and at what scale. These donations are cross-referenced against committee assignments to identify jurisdiction conflicts.
Lobbying and foreign influence networks: Registered lobbyist relationships connect corporations, foreign governments, and trade associations to their legislative targets. FARA disclosures surface foreign principals operating inside U.S. political infrastructure.
Dark money flows: IRS 990 data reveals nonprofit-to-PAC funding chains that obscure the ultimate source of political spending. Sentinel maps these 3,941 FUNDED relationships to trace money from its origin through intermediary organizations to its final destination.
Influence Exposure Scores (IES): Every legislator is scored on an eight-dimension scale integrating donation volume, lobbying access, committee power, revolving door employment history, earmark allocation, foreign interests, federal contract awards, and regulatory capture. Under IES v3.5, 1,228 legislators have been scored. The average score is 16.56; the maximum is 55.5, held by Representative Ken Calvert (R-CA). The scoring model is validated at AUC 0.9575 using an XGBoost classifier on 420-dimensional embeddings combining semantic and graph-structural features.
How data is sourced and verified
Every record in the Sentinel graph traces to a public federal filing. The ingestion pipeline runs five stages: Ingest → Resolve → Graph → Score → Verify. Entity resolution links companies, lobbying clients, and PACs that appear under different names across databases. Each ingestion event is timestamped on the XRP Ledger, creating an immutable blockchain audit trail. Use the Oracle search interface below to query the graph in plain English — results include confidence scores, supporting records, and the underlying Cypher query used to retrieve them.