Five processors, a playground to try them, managed batch for scale, and deployment options that keep your data where it belongs.
Drop in a document or pick an example, then switch between Markdown, HTML, and structured JSON.
# Invoice DX-40218
**Bill to:** Northwind Supply Co.
**Date:** 2026-05-14 **Due:** 2026-06-13
| Item | Qty | Unit | Amount |
| --------------- | --- | ------- | ---------- |
| Filtration unit | 4 | $2,150 | $8,600.00 |
| Service, on-site| 12h | $145 | $1,740.00 |
**Total due:** $18,204.50The core parsing model. It reads a document the way a person would — top to bottom, column by column — and returns clean Markdown, HTML, or JSON with the layout intact.
# Quarterly Report
| Metric | Q1 | Q2 |
| --- | --- | --- |
| Revenue | 4.1 | 4.8 |High-fidelity text recognition tuned for the hard cases: faint scans, handwriting, dense multilingual pages, and heavy mathematical notation.
{ "text": "Indenture", "conf": 0.99 }
{ "text": "of", "conf": 0.99 }
{ "text": "Grant", "conf": 0.98 }Detects regions, tables, and section hierarchy, and keeps them aligned across long documents so the structure holds together end to end.
§ 1 Introduction
├─ 1.1 Background
└─ 1.2 Scope
§ 2 Methods
└─ Figure 1Schema-driven extraction. Describe the fields you need and Docen returns them as structured data, each with a citation back to the exact source span.
{ "total_due": "$18,204.50",
"cite": { "page": 1, "span": [412, 423] },
"conf": 0.99 }An evaluation harness that scores parsing and extraction against your own labeled documents, so you can catch regressions before they reach production.
field f1 Δ
total_due 0.99 +0.01
vendor 0.97 0.00
date 0.98 +0.02Send a backlog and get structured output back, with throughput that scales to the work in front of it.
Managed batch takes a backlog of documents and runs it end to end. Throughput scales up when the queue is deep and settles back down when it drains, so you pay for the work rather than idle capacity.
Character error rate on scanned pages
Table-cell accuracy on complex layouts
Field-level F1 on schema extraction
Median processing time per page
Point your pipeline at the Docen API and keep the output format you already parse. Most teams switch one endpoint and compare results side by side.
Replace your current parser or OCR call with a single Docen API request. Same file in, cleaner structure out.
Run both in parallel and use Docen Eval to measure the difference on the documents you actually process.
Move traffic on your schedule. Keep the same output format, so nothing downstream has to change.
The fastest way to start. Send documents to the Docen API and get structured output back, with capacity that scales with your queue.
Run Docen inside your own cloud account. Documents stay within your network boundary while you use the same models and outputs.
Deploy to hardware you control, including fully offline environments. Built for teams with strict data-residency requirements.
Start on the API in minutes, or bring Docen into your own environment. Same models, same structured output, at whatever scale you run.