Skip to main content
Agentic Document Extraction (ADE) is a document intelligence platform that converts documents into reliable, structured data. Use the structured data to build retrieval-augmented generation (RAG) applications, power intelligent search systems, extract key information, and automate document processing. ADE is powered by the DPT-3 family of models, built for the era of AI agents and agentic workflows and designed around feedback from users running document-intensive workloads.

​
How ADE Works

Agentic Document Extraction provides three APIs for document processing: Parse, Extract, and Ground. Most workflows start with Parse, which converts a document into structured data. After parsing, you can use Extract to pull specific data fields from the Markdown, and Ground to map each extracted field back to its location on the page. Parse and Extract each run as synchronous requests or async jobs. Use Parse Jobs and Extract Jobs for large files and background pipelines. Ground always runs synchronously and consumes no credits.

​
Key Capabilities

  • Designed for agent systems. Built around how downstream AI agents consume document data: semantic block ids, predictable hierarchy, and Markdown ranges so you can map results back to source.
  • Hierarchical layout model. The structured output is a nested hierarchy: pages, blocks on each page, and cells within tables.
  • Fine-grained visual grounding. Bounding boxes for every block, plus per-line or per-word boxes through each block’s atomic grounding, so you can highlight or extract content at the line or word level.
  • Clean, standardized Markdown. Structured figure elements with classification labels, bracketed attestation and scan-code labels, and standardized checkboxes.

​
Important to Know

  • Parse supports PDFs, images, and Office text documents and presentations (DOCX, PPTX, ODT, RTF). Spreadsheets (XLSX, CSV) and legacy binary Office files (DOC, PPT) are not supported; for those, use Parse Gen1. See Supported File Types.
  • Password-protected files are not supported. Decrypt the file and submit an unencrypted copy.
  • Upgrading from ADE Gen1? The endpoints and response shapes changed. See the Migration Guide.

​
Find What You Need

Pick a starting point based on what you want to do.

Try It in the Playground

See Parse in action without writing code. Upload a file in your browser.

Quickstart (API)

Make your first API call: parse a document and get structured Markdown back.

How Parse Works

Understand the response shape: pages, blocks, and grounding.

Use the CLI

Parse and extract your own documents from the terminal, with results saved on your machine.