// document pipelines · Berlin
PDFs, scans, emails, faxes (yes, still). unstructred parses them, extracts structured fields and routes clean JSON to wherever your stack needs it.
Start parsing Read the quickstartyes, we know how it's spelled — the domain was cheaper.
Every pipeline runs the same battle-tested stages. You configure them in a single YAML file.
No YAML? No problem. Everything is plain Python, and the server runs the same code you test locally.
Declarative schemas mean validation comes free. Confidence scores on every field mean you can route low-confidence docs to a human, not to your users.
from unstructred import Pipeline pipe = Pipeline.from_file("invoice.yaml") for doc in pipe.watch("s3://acme-inbox/"): result = pipe.run(doc) if result.confidence > 0.9: pipe.emit(result, sink="erp") else: pipe.emit(result, sink="human-review")
A few of the 300+ teams running unstructred in production.
“We replaced a 14-person data entry outsourcing contract with one pipeline and two reviewers. The OCR on our scanned Portuguese invoices is frankly better than we expected.”
“The schema validation caught a supplier that changed their invoice layout overnight. Our old parser just silently returned garbage for a week.”
“Fax-to-JSON in production. In 2026. I don't fully understand it either, but the clinics love it.”
No sales call. No credit card. Just an API key and a schema file.
Get started Read the blog