Tabscanner extracts structured data from receipt images through an API — merchant, date, line items, tax and totals — using a layout-aware neural OCR pipeline across multiple languages and currencies, with 90% of requests returning in under two seconds. SDKs cover Python, JavaScript, C#, Node, Go, Ruby and PHP, and a human-in-the-loop option is available for higher accuracy. It is SOC 3, ISO 27001 and GDPR certified. Founded in December 2016 by Rashad Al-safar and Ben Smith, with offices in Dubai, Texas and Tokyo.
Two things put this ahead of the newer entrants around it. **It has been running since 2016**, which in a category full of six-month-old wrappers is the single most useful fact about a service you would build expense processing on top of. And the pricing is fully published per credit — free at 200 a month, $24 for 300, $360 for 6,000, with named overage rates — so the unit economics of an integration can be worked out before any conversation.
The accuracy claim needs reading with care. ‘99.99% Accurate’ is not achievable on the real input this handles: a creased thermal receipt photographed at an angle in bad light degrades any OCR system, and **the existence of the human-in-the-loop tier is itself the admission** — you do not sell human review on top of a 99.99% pipeline. Treat the automated path as very good and budget for exceptions. The dedicated Pro Service at $500 a month for data-team tuning points the same way, and is worth costing in for high-volume deployments.
