# Documentu

> Document intelligence for the MENA region. Arabic-first, API-first, and sealed inside your network when you need it. Documentu reads Arabic and English documents, traces every value to the exact place on the page it came from, and puts it to work in the systems you already run — in our cloud, or inside yours.

## What it does

- Every extracted value carries the page and the bounding box it came from, and the engine that read it.
- Arabic-first: script identification, right-to-left and mixed reading order, tables that break across pages, handwriting and stamps, and Hijri dates kept exactly as written.
- The documents that need a person arrive grouped by what went wrong, not as a queue to open one at a time. Every correction teaches the next run.
- Each page goes to the cheapest engine that can read it, and the routing decision is recorded per page.

## Benchmark

On public document-parsing benchmarks, the engine behind Documentu outscores GPT-4o, Gemini 2.5 Pro and Mistral OCR, and it is small enough to run on a single GPU. Higher is better; OmniDocBench scores are 100 × (1 − edit distance). Competitor figures are as reported in the same evaluations.

| Benchmark | Documentu | Gemini 2.5 Pro | GPT-4o | Mistral OCR |
| --- | --- | --- | --- | --- |
| OmniDocBench, overall (English) | 87.5 | 85.2 | 76.7 | 73.2 |
| OmniDocBench, tables (TEDS) | 88.6 | 85.8 | 72.0 | 75.8 |
| Multilingual parsing, 126 languages incl. Arabic | 82.3 | 74.9 | — | — |
| olmOCR-Bench, overall | 79.1 | — (Gemini 2.0 Flash: 63.8) | 69.9 | 72.0 |

Measured on a single cloud GPU: 4.4 seconds per page.

## The Black Box

Documentu sealed inside your own network, every model included. No call goes out: not to us, not to an AI provider, not to anyone.

- **Our cloud.** Managed by us; the fastest way to start.
- **Your servers.** Your hardware or a dedicated box. Your Postgres, your storage, your keys. Reachable for updates, nothing else.
- **Air-gapped.** No route to the internet, ever. Updates arrive as signed images, handed over offline.

Properties: zero calls out; every model inside; encryption under keys you hold; an audit trail of every read, not just every write; one image and one database, small enough for a security team to review. Built for banks, hospitals, government and defense.

## How it works

1. **Intake.** Through the API, an email address we give you, the mailbox you already have, or the shared drives and ERP inboxes where documents already land.
2. **Read.** Every page goes to the engine that reads it best — layout, tables, handwriting, stamps and reading order, in Arabic and English.
3. **Understand.** It knows what kind of document it is holding, and pulls the fields your business needs into your schema.
4. **Check.** Totals that must add up, suppliers that must exist, amounts that must match the purchase order. What does not pass goes to a person, with the reason.
5. **Act.** Straight into your ERP, CRM and approval flows, or out as signed webhooks and JSON, every value carrying its page and coordinates.

## Automation and integrations

The automation layer turns what Documentu read into finished work: records created, matched, approved and filed, each linked back to the page it came from.

- ERP and accounting: SAP, Oracle NetSuite, Microsoft Dynamics 365, Odoo, Sage, QuickBooks, Xero, Zoho Books.
- CRM: Salesforce, HubSpot, Zoho CRM, Dynamics 365 Sales.
- Where documents arrive: Outlook, Gmail, SharePoint, Google Drive, Dropbox, WhatsApp Business.
- Where people work: Microsoft Teams, Slack, approvals, webhooks and the API.

## What it does not do

- It does not tell you a stamp, seal or signature is genuine. It finds them and keeps them as evidence for a person to judge.
- It does not convert Hijri dates; it keeps them as written.

## API

The HTTP API is the programmatic interface, authenticated with an API key.

Signed webhooks with retry, backoff and a replayable delivery log. Sandbox mode runs a document against your workflow's own unpublished draft without it counting against your usage.

## Early access

Documentu is opening to a small group of early customers across the region. Pricing will be a licence and volume bands, never per seat and never per review. To get early access, join the waitlist at https://www.documentu-ai.com/#early-access.

## Links

- [This page as Markdown](/index.md)
- [This file](/llms.txt)