Turn complex documents into structured knowledge. Understand layouts, reading order, tables, charts, and images, all organized into the Paper Graph.
Every part of a document has a meaning,
and a place in the order you'd read it.
Quarterly Performance Report
A summary of our marketing performance and key metrics for Q2 2024.
| Channel | Visitors | Signups | Conversion |
|---|---|---|---|
| Organic | 12,840 | 2,496 | 20.0% |
| Paid | 8,320 | 1,248 | 15.0% |
| Social | 6,100 | 854 | 14.0% |
| 2,950 | 531 | 18.0% |
Every part is a node that knows what it is, in the order a person reads them.
Tables, charts and images come with summaries,
insights and what's in them.
Data is filled in once a node is enriched. Titles get their heading level, text nodes carry no extra data.
Load only the nodes you need,
filtered by type or page.
Its place in the reading order, its text,
and exactly where it sits on the page.
An ID, its type, its place in the reading order and the page it's on.
The plain text, and the entries it's made of, each with its position on the page.
Revenue increased by 24% this year.
refs[0] x 0.08 · y 0.42Every node knows where it is on the page, with a rendered image of that region to show or cite.
Move to the nodes before and after, or get the content around a node in one call.
Build viewers, extraction and RAG on top of the graph,
or let AI agents read your papers on their own.
Build viewers, extraction and RAG on the graph, and turn any search match into the passage around it.
Connect AI agents through our hosted MCP, and they search your papers on their own, answering from the exact nodes and pages that match.
Secure, compliant and built for scale.
Process one document or millions with the same API.
Your data stay yours. We never train on it.
Simple APIs, SDKs and tools to build faster.