Rev DeclassifiedFind the records they released.
One search across the National Archives, nearly a million CIA Reading Room documents, the State Department, NSA, OSS, FBI files, Project Blue Book, MKULTRA, the JFK collections, DocumentCloud, the Library of Congress and the Federal Register. Read the full text, open the PDF, decode the redactions, highlight passages, share quotes and download everything — every result links to its original record.
Built for accuracy
Live from the source
Every search goes straight to each archive's public API at the moment you search. Nothing is rewritten or invented, and every result links to its original record.
Verifiable queries
Each archive panel shows the exact query sent and the count returned, so you can repeat the search on the archive's own site.
Honest failures
If an archive is slow, rate-limited or offline you see that archive's error — never a fake "0 results".
Official links for mirrors
CIA Reading Room results come from the Internet Archive's mirror; each one also links to the matching official cia.gov page, derived from its document number.
Redactions decoded
Open any document with text and the Analysis tab finds release stamps (e.g. "Approved For Release 2005/08/22 : CIA-RDP…"), FOIA exemptions (b)(1)–(b)(9), 25X codes and classification markings, and explains each.
AI that shows its work
The optional assistant reads only the document or the result list. Every quote it gives is checked against the source text and marked verified or not found; every [n] citation is checked against your results.
Declassified collections searched
Mirrored on the Internet Archive; counts as listed by archive.org on October 22, 2026. Pick one in the Collection menu to search it alone.
Frequently asked questions
Where do the search results come from?
Live from the public APIs of the National Archives Catalog, DocumentCloud, the Internet Archive (including its mirrors of the CIA Reading Room, State Department, NSA, OSS, FBI and other declassified collections), the Library of Congress and the Federal Register. Agencies without a public search API are reached through direct search links in the sidebar.
What is the CIA Reading Room mirror, and how is it different from cia.gov?
The Internet Archive hosts a copy of close to a million CREST documents with OCR text, which can be searched by API — cia.gov's own reading room can't. Each result links to the archive.org copy and to the official cia.gov page built from its document number. Mirrors are copies uploaded to archive.org; for citation, confirm on the official page.
What do (b)(1), (b)(6), 25X1 and similar codes mean?
(b)(1)–(b)(9) are the nine Freedom of Information Act exemptions, such as (b)(1) for classified national-security information and (b)(6) for personal privacy. 25X1–25X9 are Executive Order exemptions from automatic declassification after 25 years, such as 25X1 for intelligence sources and methods. Open a document's Analysis tab to see every code found and what it means.
Does the AI make things up?
It receives only the document's own text (or your result list's catalog data) and is told to answer from that alone and never to guess at redacted text. Each quote it gives is automatically checked against the document — marked ✓ verified or ⚠ not found word-for-word — and every [n] citation is checked against the result list. Treat AI output as a reading aid and confirm in the document.
Why are some dates "undated" or only a year?
Archives record dates differently. Where an archive stores a year-only date it's shown as a year. DocumentCloud only records when a file was uploaded, so the app never shows that as the document's date; it reads the year from the title or description when one is given.
Can I highlight, share and download documents?
Yes. In any document’s Full text, select words and tap a color to highlight them, with optional notes. Share any passage as text, as a link that reopens the document at that exact passage, or as a ready-made quote image for social media. Download the PDF, the full text, page images, a research report (with your notes, highlights and the redaction analysis), or save the report as a PDF. Highlights are stored only in your browser.
Can it read documents aloud?
Yes — with the Kokoro neural voice (male American narrator) running in your browser, highlighting each passage as it reads. The voice model downloads once (about 90 MB) and is cached. If your device can't run it, the app says so and uses your device's built-in voice.
Can I install it on my phone?
Yes. Tap Get the app. Android and desktop Chrome or Edge install directly; on iPhone or iPad tap Share, then Add to Home Screen.