DeclassDB · ResilientMind AI (SDVOSB)

Every declassified archive. One search. Grounded AI.

DeclassDB unifies AI-assisted search across declassified CIA, FBI, NSA, State Department, NARA and DoD records — full text, citations, OCR'd PDFs, and entity extraction, from a single key-gated API.

Open DeclassDB.com ↗ API docs ↗

A live, public corpus

DeclassDB.com runs in the open against decades of declassified U.S. government records — the hardest public dataset available. These are measured production figures, not projections.

7 Federal agency sources (CIA · FBI · State · NSA · NARA · DoD · NS Archive)
1.05M CREST documents full-text — OCR’d and searchable inside the PDF
126,000+ Records indexed across the multi-agency corpus
< 2s Cross-agency blended search, merged & de-duplicated
$0 Per-query edge cost — client-side FTS5 search
100% Local-first storage — no accounts, no search logs

Capabilities

One call fans out to every agency, merged and de-duplicated — for researchers, journalists, OSINT analysts, and historians.

Unified cross-agency search

One query reaches every source and returns a merged, de-duplicated result set: GET /api/search?q=mkultra.

Precision filters

Scope by agency and date range — &source=cia,fbi,nara, &date_from=1960&date_to=1965 — with paging and limits.

Full-text PDF & OCR

Search inside millions of declassified PDFs down to the page, with server-rendered report export (&format=pdf).

Entity extraction

People, organizations, and locations pulled from document OCR, with AI summaries that cite the underlying records.

Semantic “more like this”

Vector search over the CREST corpus surfaces related documents that keyword search alone would miss.

Saved-search RSS

Subscribe to a query — https://declassdb.com/feed?q=… — and get new matching releases as they're indexed.

API & CLI — built for tooling

The whole API is key-gated: no key, no serve. A shared demo key is available for evaluation; Pro and Researcher keys unlock production volume.

curl -H "X-API-Key: $DECLASSDB_KEY" \
  "https://declassdb.com/api/search?q=MKULTRA%20subproject&source=cia&format=json"

Server-to-server JSON by default, with text and pdf formats and a free &summary=1 scope check to size a query before you spend quota.

declassdb CLI / SDK

The declassdb Python package (PyPI) and interactive REPL ride the same API — with local-Ollama RAG, a smart cache, and a personal PDF library that keeps your research on your machine.

Grounded by design

Results trace to real documents and citations. AI analysis summarizes what's in the corpus — it doesn't invent sources.

The public proof behind the classified system

DeclassDB is the open, public instance of the same document-intelligence engine that powers the AriaOS Classified Document & Case-Management System. What DeclassDB does in the open with declassified records, the classified system does behind the boundary — with classification, compartmentation, and access control layered on. Same architectural DNA as ValorDB, the veterans-appeals research corpus.

Built by ResilientMind AI LLC — a Service-Disabled Veteran-Owned Small Business. SDVOSB · CAGE 14JQ9 · UEI NW3SNPP7QWF4.

Search the declassified record yourself.

Open DeclassDB.com ↗ Pricing & API keys ↗