Cardinal is built for the hardest documents. Where LLMs hallucinate and miss structure, and traditional OCR tools are inaccurate, Cardinal delivers precise, structured data with bounding boxes you can trust.

Hi Product Hunt! This is Devi and Jianna, co-founders at Cardinal.

Today we’re launching the Cardinal API - the document workflow API for developers.

Cardinal is our second company together. Our first company, Leafpress, processed utility data. OCR was by far the hardest part - utility bills were very messy. We tried every tool we could think of and nothing solved our problem, so we finetuned our own VLM to process the data. That soon became Cardinal.

Unlike other mechanical OCR tools, Cardinal is more accurate. Unlike LLMs, we do not hallucinate and return high-quality bounding boxes. Cardinal gives you structured outputs that plug directly into your workflow.

Endpoints available today:

Split – classify docs, break apart multi-page files, auto-detect types

Extract – enter your schema, get a custom JSON output

Markdown – get clean, readable text that preserves layout

RAG – chunk intelligently for retrieval and embedding workflows

Developers are already using Cardinal for fraud detection, healthcare claims, government contracts, and invoice processing.

You can start building right now:

Quickstart guide or Upload a PDF and test it out!

Excited to see what you build!

-Devi

Great!!! Documents with complex structures and relatively illegible fonts can be recognized! Perhaps the Cardinal has great potential in the restoration and preservation of ancient books!

This looks super useful! Traditional OCR tools often miss key details, so having a reliable API like Cardinal for structured data sounds like a game changer.

Super cool + plug and play. Thanks Cardinal!!

Congrats on launching!