Skip to content
AI DOCUMENT DATA EXTRACTION

Extract data from any document.Straight into a spreadsheet.

PDFs, photos and spreadsheets become validated Excel in minutes: invoices, energy bills, statements, freight documents. Describe the fields once; Colunar reads every layout and records where each value came from.

20 free credits · No credit card · Delete your data anytime

Try it now, no account needed

Pick a document type, drop one file (up to 8 MB, 5 pages) or use a sample. Real extraction, real worker.

Document type

or try a sample

Demo files are deleted within 24 hours and are never used to train models.

Extracted records appear here as soon as each file is read.

Real extraction on the same engine as the product · Demo files are deleted within 24 h

PDF, PHOTO, SCAN, EXCEL AND CSV. ONE WORKFLOW.

PDFScansPhotosExcelCSVLegacy XLS
Built in BrazilData encrypted on AWSLGPD compliantDelete files anytimeNo card to start

01 / WHO IT IS FOR

Built for teams that receive documents in volume.

Every area has its documents and its fields. Pick yours and start from a ready-made playbook.

02 / HOW IT WORKS

From PDF to spreadsheet in three steps.

You decide what matters. Colunar finds, validates and organizes the fields, file by file.

  1. 01

    Describe the fields you want

    Pick a ready-made playbook or list your columns: supplier, date, amount, consumption. Types and required fields become a schema.

    supplierdatetotal
  2. 02

    Drop the files, in bulk

    Native or scanned PDFs, phone photos, XLSX and CSV. Each file shows up in the table as soon as it is read, with warnings when something could not be read.

    PDFJPGXLSX
  3. 03

    Check the source and export

    Every value points to the page, sheet or row it came from. Download as Excel, CSV, JSON or Markdown.

    ExcelCSVJSONMarkdown

03 / THE PRODUCT

The table fills in while the files are being read.

No hidden queue: each file shows its state, its warnings and its records as it finishes.

Colunar screen with a run in progress: file list with status and the results table being filled
A run with energy bills: files, status and the live table.
Colunar playbook editor with fields, types, required flags and reading instructions
Playbook: fields, types, instructions and reusable skills.

04 / PLAYBOOKS + SKILLS

Reading rules that apply to every following file.

A playbook defines the fields. Skills add instructions and reviewed examples that guide interpretation. See a rule in action.

Provide a document and its expected result to guide future readings.

Playbook / Finance2 skills
Normalize dates and amounts

Read dates as day/month/year. In Brazilian amounts, a dot separates thousands and a comma separates cents.

Reviewed example

Provide a document and its expected result to guide future readings.

IN THE DOCUMENTIN THE RESULT
05/09/20262026-09-05
R$ 1.250,001250.00

Skill applied

Illustrative example. Skills guide reading; review results from your own documents.

05 / CAPABILITIES

Built for documents that do not follow a template.

Stacked tables, poor scans and fields that move around are the common case, not the exception.

CSV · XLS · XLSX

Several tables in one sheet

Stacked headers, total rows and irregular regions are detected and mapped; every row becomes a record.

PDF · PNG · JPG

Scans and phone photos

Pages without selectable text are read by vision models, with enlarged regions for small print.

SCHEMA + SOURCE

Validation and source on every record

Types and required fields are checked against the schema. Issues become warnings per file; nothing is dropped silently.

Documents that usually come in

  • Invoices and fiscal notes
  • Electricity and water bills
  • Bank statements
  • Receipts and expenses
  • Purchase orders
  • Freight documents
  • Lab reports
  • ESG and emissions data

06 / COMPARISON

Why not paste it into a chatbot or use OCR?

For one document, any tool will do. For fifty a week, what matters is bulk, validation and traceability.

Manual typingTraditional OCRGeneric AI chatColunar
Fixed fields, right formatDepends on the personLoose text onlyChanges every answerValidated schema
Batches of dozens of filesHoursYes, without fieldsOne at a timeYes, with a live table
Source of every valueNoNoNoPage, sheet and row
Reusable rulesIn someone's headTemplate per layoutRewrite the promptPlaybooks and skills
ExportIt is the spreadsheetTextCopy and pasteExcel, CSV, JSON, Markdown

07 / PLANS

Pay only for what you process

Start with 20 free credits. Pick a monthly plan, top up when you need, or talk to us for volume.

Starter

For individuals and first workflows.

R$30/month

  • 200 credits / month
  • +100 credits in your first month

Pro

Most popular

For teams processing documents every week.

R$180/month

  • 1,500 credits / month

Business

For operations at scale.

R$600/month

  • 6,000 credits / month

Enterprise

Custom plan, pay-as-you-go or invoicing for your volume.

Credits per file

  • CSV1 credit
  • XLSX2 credits
  • XLS2 credits
  • PDF4 credits
  • PNG3 credits
  • JPG3 credits
  • JPEG3 credits

Top-up

100 credits that never expire · R$20

Plan credits reset every month.

BEFORE YOU START

Frequently asked questions

Which files can I upload?

Native or scanned PDFs, PNG, JPG and JPEG images (including phone photos) and XLSX, XLS and CSV spreadsheets, up to 50 MB per file and 200 pages per PDF.

Do I need to build a template for each layout?

No. You describe the fields once, in a playbook. Colunar reads each document by content, not by position, and adapts to bills, invoices and statements from different issuers.

Does it work with energy bills from any utility?

Yes. Utility, consumer unit, reading period and consumption are read regardless of layout. Specific rules, such as using the reading period instead of the billing month, go in as playbook instructions.

Does it read scans and phone photos?

Yes. Pages without selectable text are read by vision models, with enlarged regions for small print. Whatever is unreadable becomes a warning instead of an invented value.

How accurate is the extraction?

Every record is validated against the schema you defined, and issues are reported per file and page. Accuracy depends on the document and your rules; that is why the source of every value is recorded for review, and reviewed examples (skills) guide the following readings.

How do credits work?

Each file costs credits by type: CSV 1, XLSX and XLS 2, image 3, PDF 4. Credits are charged when processing starts and returned if the file cannot be processed.

What happens to plan credits at the end of the month?

Plan credits are valid until the next renewal and do not roll over. Welcome and top-up credits never expire.

Are my documents private?

Files are stored encrypted on AWS, used only to process your extractions and never to train models. You can delete any file, extraction or your whole account at any time.

What export formats are available?

Excel (XLSX), CSV, JSON and Markdown, for a single run with all of its files. You can also download the original files.

Do I need to install anything?

No. It is a web service: sign in with your email or Google, drop the files and download the result in the browser.

Is there an API or integration with other systems?

Not yet. Today the export is by file (Excel, CSV, JSON). If you need an integration, talk to us about the Enterprise plan.

How do I pay and which payment methods are accepted?

Payments are processed by Stripe with a credit card. Monthly plans renew automatically and can be canceled at any time; top-ups are one-time purchases. Every purchase comes with an invoice.

LESS TYPING. MORE DATA.

Your next batch of documents can come out as a spreadsheet.

Create an account with your email or Google and get 20 free credits.

Start free