Classify

The Document Router for AI Pipelines

Define your categories in plain language. Classify returns the best match with confidence scores, no training data required.

Try for free Request a demo

Helping everyone from startups to Fortune 10 enterprises unlock their data.

Classify

Route documents by type before processing

Definition: Classify labels a document against a list of categories you define. Each category has a name and a list of natural-language criteria. Classify returns the best match plus a per-category and per-criterion confidence breakdown (enough signal to build threshold-based routing on top).

Who it's for: Teams that need to route different documents to different workflows. Common uses include document triage at upload, schema routing for extraction, and conditional processing settings.

The problem it solves: When users upload mixed document types, each one needs different processing. Classify sorts the pile so every document is routed to the right pipeline automatically.

Classify in the platform

How Classify connects to the rest of the platform

Endpoint

Use when

Output

How it works with Classify

/parseParse

Structured content from any document is needed for LLM or RAG use.

Structured chunks with typed blocks, bounding boxes, and confidence scores.

When routing isn't needed and one config fits all files.

/extractExtract

The fields to pull are defined and typed JSON is needed.

Schema-typed JSON with optional citations on every value.

Pulls field values. Classify is faster for routing only.

/splitSplit

One file contains multiple logical documents or sections.

Page ranges for each section, with confidence scores.

Maps sections within a file, not the file type.

/classifyClassify

Files need to be routed by type before processing.

Best-matching category with per-criterion confidence.

Read the Classify docs

/editEdit

A PDF form needs filling or a DOCX needs updating.

A downloadable edited file, plus a reusable form schema.

Produces a filled document. Classify picks which instructions to use.

Try Classify on your own documents. Open it in Studio.

Open Studio Request a demo

Where AI teams ship Classify

Triage, routing, and conditional pipelines

When your pipeline needs to handle different document types differently, Classify makes the routing decision.

Document triage at upload

Sort IDs, statements, contracts, and voided checks into the right processing queue as soon as they're uploaded.

Schema routing for extraction

Passports get one Extract schema, immigration forms get another, W-2s get a third. Classify decides which schema your pipeline calls.

Conditional Parse settings

Detect handwritten notes vs. printed forms, then apply the right Parse settings automatically.

Pipeline branching

Decide which endpoints to call based on a single Classify result at the top of the pipeline.

Confidence-based human review

Auto-route documents above your confidence threshold. Flag low-confidence results for review.

Catch-all and ambiguity detection

Add an “other” category for anything unexpected, or compare top scores to spot ambiguous documents.

Try Classify on your own documents. Open it in Studio.

Open Studio Request a demo

Why Classify

Why teams switch to Classify

01
You define the categories
You write the categories and the criteria that distinguish them. Classify matches documents against what you define.
02
Per-criterion confidence on every category
Each criterion is scored high/low; each category gets a 0–1 score from its match rate. Not a black box.
03
Optimized for inline latency
Synchronous only. Fast enough to run at the top of a pipeline since it skips full parsing.
04
Threshold-based routing built in
Auto-route above 0.7, flag below it, detect ambiguity when two categories tie. Branching falls out of the score.
05
Classify beyond document type
Categories aren't limited to document types. Detect content characteristics like handwriting, tables, watermarks, or complexity to route documents to the right settings.
06
Schema refinement loop
Low-confidence results show which criteria to improve. Adjust and re-run until routing is reliable.

How Classify works

How Classify works in three steps

STEP 01
Send file + classification_schema
Upload a file or point at a URL. List your categories with their criteria.
POST /classify
STEP 02
Lightweight ingestion
Classify reads the document for high-level understanding without a full Parse. Each criterion is evaluated as a yes/no signal.
no full parse
STEP 03
Best match + structured reasoning
You get the best-matching category plus a confidence score and per-criterion breakdown for every category.
result.category