Document Intelligence

Hush Cyclops: deterministic document intelligence where your documents never leave the building.

Extract text from any document on hardware we own. We don't save your documents, they are processed, then secure-wiped. No LLM touches your data during text extraction. Deterministic, auditable, sovereign.

Available via REST API, MCP protocol, and right here.
We own the hardware
Our own hardware, not Big Tech
Full audit trail
ICO ZC126901

Use Cyclops right here.

Upload a document. Choose a mode. Get results instantly.

Drop a file here or click to browse
Max 50MB
PDF DOCX JPEG PNG WebP TIFF
Processing...

Sign in to use Cyclops.

Sign In

Two engines. Completely isolated.

Text extraction and visual description are separate paths. OCR output never touches the vision model.

OCR Engine

Direct text extraction for text-based PDFs. On-device OCR for scanned documents and images. Native parsing for Word documents. Every character comes directly from the source. No LLM involved.

Deterministic text extraction

Vision Engine

An on-device vision model for describing diagrams, photos, screenshots, and charts. The only path that uses an LLM. Completely separate from OCR.

Visual understanding
Key point: Text extraction and visual description are completely isolated. OCR output never touches the vision model.

Every document format your organisation uses.

From single-page receipts to 200-page contracts.

PDF (up to 200 pages) DOCX JPEG PNG WebP TIFF

Cloud OCR sends your documents abroad. Ours does not.

AWS Textract, Google Vision, and Azure Document Intelligence are operated by US companies. The CLOUD Act gives US authorities the legal power to compel access to your documents.

Cloud OCR Providers

  • Data leaves UK to Big Tech infrastructure
  • Subject to CLOUD Act (H.R.4943)
  • Per-page fees that scale unpredictably
  • Vendor lock-in with proprietary formats
  • No visibility into how documents are processed

Hush Cyclops

  • All processing on hardware we own
  • Structurally outside the CLOUD Act
  • Flat-rate pricing, no per-page fees
  • Open standards, no vendor lock-in
  • Full audit trail for every document processed

Document intelligence for every industry.

From invoices to blueprints. Every document processed on our own hardware.

Invoice and receipt processing

Extract line items, totals, VAT, and supplier details from invoices and receipts for automated accounting workflows.

Contract text extraction

Pull clauses, dates, parties, and obligations from contracts for legal review. Handle multi-column layouts and annexes.

Medical record digitisation

Convert paper medical records, prescriptions, and clinical notes into structured digital text for NHS systems.

Insurance claim processing

Extract claim details, policy numbers, and supporting evidence from scanned claim documents and photographs.

Academic paper analysis

Process research papers, extract citations, tables, figures, and structured data from complex academic layouts.

Government form processing

Digitise and extract data from government forms, applications, and correspondence at scale.

Architectural drawing analysis

Interpret blueprints, technical drawings, and architectural plans. Extract annotations, dimensions, and labels.

Handwritten note transcription

Convert handwritten notes, letters, and annotations into searchable, editable digital text using vision AI.

REST API. Bearer token auth. Three modes.

Integrate Cyclops into your existing workflows in minutes.

Extract Text (OCR)
curl -X POST https://api.hush-ai.uk/v1/cyclops/process \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -F "[email protected]" \
  -F "mode=ocr"

# Response
{
  "text": "extracted content...",
  "mode": "ocr",
  "filename": "invoice.pdf",
  "file_type": "application/pdf",
  "chars": 4820
}
Describe Image (Vision)
curl -X POST https://api.hush-ai.uk/v1/cyclops/process \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -F "[email protected]" \
  -F "mode=eye" \
  -F "instructions=What is the total shown in this chart?"
Structured Extraction (JSON)
curl -X POST https://api.hush-ai.uk/v1/cyclops/process \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -F "[email protected]" \
  -F "mode=extract" \
  -F 'template={"vendor": "string", "total": "number", "date": "date"}'

# Response
{
  "text": "{\"vendor\": \"Acme Corp\", \"total\": 1530.0, \"date\": \"2026-03-15\"}",
  "mode": "extract",
  "chars": 58
}

Enterprise-grade security. No compromises.

Every layer hardened. Every request audited.

🔐

Bearer Token Auth

Every request authenticated with scoped bearer tokens. No anonymous access.

🛡️

Path Guards

Strict path validation prevents directory traversal and unauthorised file access.

⏱️

Rate Limiting

Per-token rate limiting prevents abuse and ensures fair resource allocation.

📜

Audit Logging

Every document processed is logged with timestamp, token, and metadata. Exportable.

🔒

Our Hardware, Our Rules

Documents are processed on hardware we own, then wiped. No cloud, no third party, we chose to keep no conversations or documents.

🔒

ICO Registered

Data controller ZC126901. Built to GDPR & DPA 2018. DPIA drafted; signed DPA executed before any live personal data flows.