Native-text PDF
Direct text extraction from machine-generated PDFs — the fast path, with structure and coordinates preserved for downstream classifiers.
Built by Crescentic LLC
The document spine for regulated workflows.
docABL ingests, classifies, extracts, and validates the documents your regulated process depends on — with full lineage, tenant isolation, and an audit trail that holds up.
Row-level security Multi-tenant Lineage-first
Three-lane parse front
One intake. Three engines. No documents left behind.
Every submission is routed to the lane that will actually read it. The choice is recorded, so the audit trail tells you not just what was extracted but how.
Image OCR
Scanned and photographed documents pass through an OCR pipeline tuned for regulated-form vocabulary, with confidence scoring per region.
Image-PDF rasterizer
Hybrid PDFs whose pages are really images get rasterized server-side and routed back through OCR — no silent data loss.
Tenants & accounts
Isolation that's enforced where it counts.
Every document, classification, and validation issue belongs to a tenant. Postgres row-level security — not application code — decides who can see what. Personas (service, reviewer, operator) layer on top without bending the model.
Tenant
The boundary. Your org's data never crosses it.
Accounts
Real people, mapped to auth.uid via a server-side trigger.
document_access
Per-document grants for cross-team reviews.
Ready when your documents are.
Sign in to start submitting documents. Uploads and the document workspace light up next.