Introducing Drex for Document Processing 📑 Decision-powered parsing that catches errors before they reach your agents. → #1 on our parsing leaderboard → Beats GPT-6 Astra, Claude Opus 4.7 & every parser tested → 38% faster corrections on RealDocBench Try it:

#1 of 13 on our parsing leaderboard 👇

Why this matters for agents: your agent acts on whatever the parser gives it. One wrong number in a table and it approves the wrong invoice. Drex checks every block against the page before your agent sees it, and if a fix fails, it keeps the original reading, so you never get a blank page.

Use it in your document workflows, from invoice processing to financial reports and AI agents. → 30 file types: PDF, Word, Excel, Parquet, email, images & audio → Our cloud, your AWS/Azure, or fully on-prem → Console, REST API, CLI, Python & TS SDKs → Same API key as Drex

Tech Blog Post:

This is a pretty impressive upgrade for document processing.

Huge for production workflow and enterprises. everywhere in my workflow, it's just document parsing after document parsing. rooting for you guys. 👏

Worth adding meraGPT Decider to the RealDocBench runs. It checks extracted fields as typed yes/no or choice questions with a probability each, so low-confidence fields go to review. KL 0.116 on typed decisions, $0.04 per million input tokens, output free.

Accuracy matters more than speed

Catching errors before agents see them is huge

Used the application. It's actually very awesome

When a correction fails and Drex keeps the original reading, does that block get marked for review? I'd want the invoice workflow to know which amount still needs a pair of eyes.

what about hidden texts into documents? is it only OCR based or the older parsing methods are used too for docs?