@pipeworx/crs-reports

Connect: https://gateway.pipeworx.io/crs-reports/mcp · Install: one-click buttons

Tools: 4

Congressional Research Service reports — Congress’s own nonpartisan analysis of statutes, federal programmes and policy options — searchable by subject and readable in full, with each report’s authority, citations, licence and age stated on every response.

Tools

• crs_search(query?, topic?, product_class?, author?, published_since?, published_before?, limit?) — ranked search over report titles and the summaries CRS publishes with them. Returns titles, report numbers, CRS subject topics, authors, publication dates and a summary snippet. Ranks titles and summaries, not report bodies — that boundary is stated in every response as searched, because a miss otherwise reads as “CRS has never written about this”.
• crs_report(report_id, include_text?, max_chars?) — one report in full by its number (R43255, IF13006, LSB10123). Returns the complete current-version text plus authors, topics, the bills and public laws it cites, and its age.
• crs_recent(days?, topic?, product_class?, limit?) — what CRS has published or revised lately, newest first. The change feed, for tracking the analysis that follows a bill or a rule.
• crs_topics(facet?, limit?) — CRS’s own subject taxonomy with per-term counts, plus analysts and product families. Call it before filtering: CRS writes U.S. Trade Policy, and a guessed Trade matches nothing.

Auth

Keyless.

Data sources

• https://api.congress.gov/v3/crsreport — the Library of Congress’s record for each report: title, publication date, current version, authors, CRS subject topics, related bills and public laws, and CRS’s own summary. api.data.gov fronted.
• https://www.congress.gov/crs_external_products/ — the report PDFs the API points at; the text comes from the current version of each.

What the data does and does not say

CRS analysis is persuasive, never binding. It is not law, it binds no court and no agency, and it is cited as background and legislative context. Every response carries authority.binding_status: "persuasive" and a binding_note saying this in words, so a downstream agent cannot mistake a CRS explanation of a statute for the statute.

The set is the currently-active one, 2018 to today — 14,115 reports at the time of writing, which is what CRS presently lists as active rather than the full historical archive. A report withdrawn or superseded by CRS leaves the set.

Licence: none needed. CRS reports are works of the US federal government and 17 U.S.C. §105 puts them outside copyright, so license_terms.encumbered is false and no attribution is required. The one caveat, which CRS prints itself and this pack repeats in license_caveat: an individual report may reproduce a third-party chart or photograph, and §105 does not reach those.

Traps, measured 2026-09-15

Worth knowing before touching the ingest (scripts/ingest-crs-reports.mjs):

• sort is accepted and ignored. sort=updateDate+asc, sort=updateDate+desc and a nonsense parameter all return the same first row. The list is always newest-first by updateDate. Nothing errors, so ordering built on sort is wrong and looks right.
• fromDateTime / toDateTime do work, and they filter on updateDate — 14,115 unfiltered against 585 for fromDateTime=2026-09-01. That pair is the change feed that keeps the daily pass incremental.
• updateDate is an America/New_York wall time wearing a Z. Fourteen sampled reports each had a PDF Last-Modified exactly 4h (EDT) or 5h (EST) after their updateDate, to the second; the DST correlation is what rules out a fixed publishing lag. The PDF’s header is the same event in real UTC and is what gets stored.
• One walk of the list does not see the whole corpus. It returns exactly the 14,115 rows it claims and 14,097 distinct ids: 18 come back twice, every duplicate pair sharing one updateDate (2025-09-08T11:20:24Z). That is an unstable tiebreak — ordering is by updateDate alone and thousands of rows share a stamp from one bulk republication — so about as many reports are skipped as repeated while the count looks perfect. The ingest collapses duplicates and walks again.
• The list row lags its own detail record. Seconds apart for IF12988: list.version 1 / list.updateDate 22:09:04 against currentVersion 6 / updateDate 22:38:57. Enumerate from the list; take every value from the detail record.
• The HTML rendition is unreachable and the PDF is fine. The /crs_external_products/<T>/HTML/<id>.html path answers 403 with Cf-Mitigated: challenge from Worker egress and from a laptop, with and without a full browser header set. The PDF path answers 200 from both. A bare HEAD on the PDF gets 403 where GET gets 200, so a HEAD-only probe condemns the working path.
• summary is the whole document for short products and only the summary for long ones. R43255 is 3,754 characters of summary against 61,691 of report. text_source on every response says which rendition the text came from.

Tools

• crs_search — Find Congressional Research Service (CRS) reports on a policy topic, statute or programme
• crs_report — Read one Congressional Research Service report in full by its report number (R43255,
• crs_recent — What the Congressional Research Service has published or revised lately — the newest CRS
• crs_topics — The Congressional Research Service’s own subject taxonomy, with how many reports carry