capturall

Everything to turn documents into data

Capturall combines strong OCR with AI extraction and shows where every result came from. So you check instead of retype.

Join the waitlist
An employee compares a paper document with the screen

Read

Every page as text, blocks and positions.

Read turns PDFs, scans and phone photos into structured text. Every block gets a type (title, text, table, signature, stamp), a position on the page and a confidence score. Tables stay tables, even when they run across two pages.

  • Works on scans, photos and handwriting
  • Output as Markdown, JSON or HTML
  • 1 credit per page

delivery_8812.pdf · page 1

5 / 5

Delivery note 8812

99%

Van Dijk Logistiek B.V., Havenweg 12, Moerdijk

98%
ItemQty
Euro pallet, wood 120×8022
Stretch film, roll6
Corner board, cardboard88
96%
91%
Received
88%

Split

One bundle in, separate documents out.

A driver scans ten pages in one go: CMRs, delivery notes and a weighbridge ticket mixed together. Split finds where each document starts and ends, based on categories you describe yourself, and hands each part to the right flow.

  • Your own categories in plain language
  • Page range and reason for every part
  • 2 credits per page, Read included

scan_bundle.pdf · 10 pages

3
CMR consignment notep. 1–4
Delivery notep. 5–8
Weighbridge ticketp. 9–10

Extract

Your schema, filled in, with sources.

You describe the fields you want, or let Capturall suggest a schema from an example. Extract fills in the schema, including nested lists such as invoice lines, and keeps the page and spot of every value.

Schema → Result

6 / 6
invoice_number
F-2026-0418
invoice_date
2026-11-03
supplier.vat_id
NL859310572B01
lines[0].amount
1.240,00
lines[1].amount
86,50
total_incl_vat
1.605,07
  • Nested fields, lists and your own types
  • Source and confidence for every field
  • 2 credits per page, Read included

Classify

Know what arrives, and route it.

Classify decides the document type with a confidence you can rely on. Above your threshold the document goes straight to the right flow; below it, to a person. An invoice never ends up in the CMR flow.

  • Your own types and thresholds
  • Doubtful cases go to the review queue
  • 0.5 credit per page, at least 1 per document

Incoming document

scan_0042.jpg
CMR consignment note97%
Invoice2%
Delivery note1%

To flow: Transport CMR

Fill

Fill in forms from your data.

Fill works the other way round: you provide an empty form and your data, and Capturall fills it in. Every filled field is marked, so you see at a glance what was set automatically and what is still missing.

  • PDF forms and scanned forms
  • Filled fields marked
  • 5 credits per page, 1 when all data is already there

Employer statement

3 / 4

Employee name

J. de Vries

Employed since

01-03-2021

Contract type

Permanent

Gross annual salary

Still to fill in
Employer signature

Filled automatically

Flows

Steps in a row, tested and versioned.

A flow is a chain such as Classify, Split, Extract and a webhook. You build it in the Studio, test it on your own examples and put a fixed version live. New versions are compared against a test set first, so a change never breaks something quietly.

Transport CMR

Live: version 3
  1. 1Classify
  2. 2Split
  3. 3Extract
  4. 4Webhook to TMS

Test set: 48 documents, 97.9% of fields correct

  • Versions with their own ID
  • Evaluate: accuracy per field and version
  • Roll back to a previous version in one click

Ask

Ask a document questions.

Ask answers questions about one document or a whole folder, pointing to the page where the answer is. In the Studio, Ask also helps you build: have it suggest a schema or change a flow in plain language.

  • Answers with page references
  • Change schemas and flows in plain language
  • 1 credit per message

cmr_20261103.pdf

1 credit
You: When was the load taken over, and by whom?

On 3 November 2026 in Moerdijk, by Van Dijk Logistiek B.V.

Source: page 1, box 4 and box 16

Ask a question about this document

API

Everything you build in the Studio, over REST.

Every flow has its own endpoint. You send a document or a link, Capturall processes it in the background and reports the result through a webhook. API keys per environment, logs per call and an OpenAPI specification.

curl https://api.capturall.com/v1/flows/fl_transport_cmr/runs \
  -H "Authorization: Bearer $CAPTURALL_KEY" \
  -F file=@cmr_20261103.pdf

→ 202 { "run_id": "run_8kq2", "status": "queued" }

POST https://tms.example.com/hooks/capturall
  { "run_id": "run_8kq2", "status": "done",
    "fields": { "sender": "Van Dijk Logistiek B.V.",
                "place_taken_over": "Moerdijk" } }

Result via webhook

  • Signed webhooks
  • API keys per environment, stored only as hashes
  • OpenAPI specification and logs per call
A phone captures a crumpled delivery note on the dashboard, four corners in view
Wrapped pallets with a delivery note in a document pouch
Hands sorting a stack of invoices and receipts on a desk

A citation on every field

Click a field and the document jumps to where the value is. Fields with low confidence go to a review queue, so a person only looks where it matters.

Fields

Reading, 0 of 5

Pick a field to see its source

The Studio

Build flows visually, test on your own documents and compare versions with test sets before anything goes live.

  • Schema builder with nested fields, lists and an AI suggestion
  • Results as formatted, JSON, Markdown or table
  • Evaluate: measure accuracy per field and per version
  • Deploy as a fixed version with its own ID

Documents arrive the way you want

  • Upload in the web app or from mobile (photo)
  • Forward by email to a dedicated address per flow
  • Pick up automatically from S3 or a OneDrive folder
  • Through the API from your own systems
Sign up for early access. It takes half a minute.Join