The web interface requires almost no learning.
Upload a file, define the fields, and click Extract. The results appear directly alongside the document. It took around two minutes to go from opening the page to receiving the first JSON output.
I first tested a standard machine-printed invoice and extracted:
“Invoice number, date, total amount, and supplier name.”
The tool performed relatively consistently with this kind of well-structured document. It found the main fields, and the numbers did not show any obvious misalignment.
However, this result should not be assumed for every PDF.
Recognition becomes less stable when scans are poor, fonts are blurry, stamps cover the main text, or handwriting is mixed into the document.
It is better suited to reasonably structured business documents than to rescuing every difficult-to-read scan.
The way fields are defined also affects the result.
For example, if you define only total_amount, the model may need to decide between a subtotal, tax amount, and final total when all three appear in the document.
Changing the description to:
“the final total amount including all taxes”
makes the intended result clearer.
The field name is therefore only the first layer. When the data could be ambiguous, it is best to add a short description.
The free allowance works well for this kind of testing.
Twenty pages can be processed without entering a credit card, allowing you to test several real file types together: ordinary PDFs, scanned documents, invoices from different suppliers, or contracts with different layouts.
This provides a much more realistic view of accuracy than uploading only one perfectly clean sample.
Pros
- Simple setup: There is no need to write regular expressions or create a parsing template for every layout.
- Custom field definitions: The same tool can be used for invoices, contracts, résumés, and other document types.
- Supports images: It is not limited to native PDFs.
- Direct JSON and CSV output: Convenient for both developers and spreadsheet users.
- No card required for the free allowance: Real documents can be tested before deciding whether to pay.
- Provides an API: Easier to integrate into batch workflows than a web-only extraction tool.
Cons
- Less reliable with complex scans: Handwriting, blur, and obstructions all affect the result.
- 20 MB file limit: High-resolution scanned contracts may need to be compressed or divided first.
- Page-based pricing: Monthly costs need to be estimated in advance when document lengths vary considerably.
- Unused allowances do not roll over.
- No refunds: Representative files should be tested before purchasing.
Best for / Not ideal for
Best for
- Finance teams: Extract fixed fields in batches from invoices, bills, and bank statements.
- HR teams: Convert names, contact information, experience, and other résumé details into structured data.
- People who regularly process contracts: Extract dates, amounts, parties, and other information from large numbers of similar files.
- Small businesses and individuals: Start from the web interface without building an OCR and parsing workflow.
- Developers: Convert PDF data into JSON for databases, automation, or AI workflows.
Not ideal for
- People primarily processing handwritten documents: Results are highly affected by handwriting and scan quality.
- Severely blurred or distorted documents: If the source image quality is too poor, field extraction will also be unreliable.
- Systems requiring completely real-time results: Each parsing request still requires some processing time.
- Teams whose data cannot leave the internal network: Files must be uploaded for cloud processing, making it unsuitable for documents that cannot be transmitted externally.
- Organizations requiring on-premises deployment: The current product is primarily offered as SaaS and an API.
Comments (0)