Unstructured
by Unstructured Technologies
ETL for unstructured data: turn PDFs, tables, emails and scans into AI-ready JSON for RAG and agents
Unstructured is a document ETL platform that converts 65+ file types — PDFs, tables, slides, emails, scans — into structured, chunked, embedded data for RAG and AI agents. It is for enterprise data and AI platform teams that need compliant (FedRAMP High, HIPAA, SOC 2) ingestion pipelines instead of hand-built parsing code.
Unstructured is an AI data-preprocessing company founded in 2022 by CEO Brian Raymond, a former CIA analyst, that sells an ETL layer for turning messy enterprise documents into LLM-ready data. It ships in two forms: an Apache 2.0 open-source Python library (about 15.5k GitHub stars) that partitions documents into typed elements such as titles, narrative text and tables, and a commercial Unstructured Platform that adds higher-accuracy table and OCR extraction, chunking, enrichment, embedding generation and source/destination connectors for cloud storage, SharePoint-style repositories and vector databases, run as no-code workflows or via API. The platform handles 65+ file types, and its hosted API launched in January 2024 with an enterprise platform following. Unstructured raised a $25M Series A in July 2023 and a $40M Series B in March 2024 led by Menlo Ventures with Databricks Ventures, IBM Ventures and NVIDIA's NVentures, bringing total funding to about $65M; at that time it cited 1,000+ paying API customers and use at over a third of the Fortune 500. IBM announced an OEM partnership in October 2025 to embed Unstructured in watsonx.data, and a June 2026 release expanded its Microsoft Azure integration. The vendor lists FedRAMP High, SOC 2 Type II, HIPAA, GDPR and ISO 27001, with dedicated, in-VPC and bare-metal deployment for Business customers. Independent comparisons (Nutrient, September 2026) position it as a heuristic layout engine that gives teams control and self-hosting, while VLM-based parsers such as Reducto and LlamaParse lead on complex tables and scans — a gap the vendor acknowledges for its open-source tier.
Enterprise data-platform and AI engineering leads in regulated or government environments who must feed RAG and agents from large, mixed document estates.
Replaces bespoke parsing, chunking and embedding code with one compliant pipeline that runs in SaaS, your VPC or on bare metal.
At a Glance
- Category
- Data & Analytics
- Pricing
- Usage-based, Freemium, Contact for pricing
- Target Market
- CIOs, CTOs, Data Engineers, Enterprise Developers
- Deployment
- Cloud-first, Open-source, Self-hosted, API-based
- Founded
- 2022
- Customers
- 1,000+ paying API customers (February 2024)
Key Features
- ✓Partitioning across 65+ file types
Splits PDFs, Office files, HTML, emails and images into typed elements, preserving structure that naive text extraction destroys.
- ✓Table and OCR extraction
Platform tier preserves rows, headers and multi-page tables and handles scans and handwriting, where most RAG answers fail.
- ✓Chunking, enrichment and embeddings
Built-in chunking strategies, metadata enrichment and embedding generation produce vector-ready output without extra glue code.
- ✓Connectors and workflows
Source and destination connectors for cloud storage and vector databases run as scheduled no-code workflows or via API.
- ✓Flexible, compliant deployment
SaaS, dedicated instance, in-VPC or bare metal, backed by FedRAMP High, SOC 2 Type II, HIPAA, GDPR and ISO 27001.
- ✓Open-source library
Apache 2.0 Python library for local, air-gapped processing that shares element schemas with the commercial platform.
Capabilities
Use Cases
- •Enterprise knowledge RAG
IT teams ingest policies, contracts and slide decks into a vector store so assistants cite accurate passages instead of hallucinating.
- •Government document processing
Agencies process sensitive records inside FedRAMP High or in-VPC deployments, keeping data within their own infrastructure boundary.
- •Financial document extraction
Banks convert filings and reports with complex tables into structured JSON for analyst copilots and downstream agents.
- •Healthcare records preparation
Providers prepare clinical documents for AI search under HIPAA with zero data retention, avoiding custom parsing projects.
Ideal For
Best For
- ✓Building RAG ingestion pipelines over large, mixed-format document repositories
- ✓Regulated and public-sector workloads needing FedRAMP High, HIPAA or in-VPC processing
- ✓Teams wanting an open-source parser to self-host with a commercial upgrade path
- ✓Continuous syncing of document sources into vector databases with chunking and embeddings handled
- ✓IBM watsonx.data and Azure shops using the vendor's OEM and Azure integrations
Not Ideal For
- ✗Workloads dominated by scanned documents, complex multi-page tables or multicolumn layouts — independent comparisons find VLM parsers like Reducto and LlamaParse more accurate, and the open-source tier relies on Tesseract OCR
- ✗Teams wanting a lightweight local install — the open-source library pulls heavy dependencies such as LibreOffice and Poppler
- ✗Small projects with only clean digital PDFs, where free libraries like Docling may be enough without a paid platform
Integrations
Deployment
Market & Ratings
1,000+ paying API customers (February 2024)
Market Analysis
Pros
- ✓Broad format coverage (65+ types) beyond PDFs, praised by practitioners on HN
- ✓Strong compliance and deployment options including FedRAMP High and in-VPC/bare metal
- ✓Simple published per-page pricing with a generous 10,000-page free start
- ✓Open-source library gives an exit path and supports air-gapped processing
Cons
- ✗Open-source tier is materially less accurate on tables and scans than the paid platform, by the vendor's own admission
- ✗Self-hosted install is cumbersome, with heavy system dependencies like LibreOffice and Poppler
- ✗Heuristic layout approach trails VLM-based parsers such as Reducto and LlamaParse on complex layouts in independent comparisons
- ✗Multi-user accounts and private deployment require a custom-priced Business contract
Pricing
Open source
$0
- ✓Apache 2.0 library
- ✓Self-hosted
- ✓Community support
Free
$0
- ✓10,000 free pages to start
- ✓No card required
- ✓All features
Pay-As-You-Go
$0.015/page
- ✓After the first 10,000 pages
- ✓All features
- ✓SaaS hosting
Business
Contact for pricing
- ✓Multi-user accounts
- ✓Dedicated instance, in-VPC or bare metal
- ✓Personal support
The platform is metered per page: 10,000 free pages, then $0.015 per page on self-serve SaaS. Multi-user accounts and private deployment (dedicated, in-VPC, bare metal) require the custom-priced Business plan. The open-source library is free but, per the vendor, less accurate on tables and documents.
Security & Compliance
Connect
Sources
This page was written from 7 sources, 4 on domains other than unstructured.io.
- 1.unstructured.io — pricingvendor
- 2.unstructured.io — unstructured.iovendor
- 3.unstructured.io — pressvendor
- 4.github.com — unstructured
- 5.siliconangle.com — ai focused big data startup unstructured raises 40m make dat
- 6.nutrient.io — best document parser llamaparse unstructured reducto
- 7.hn.algolia.com — search
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Tacnode Context Lake
PostgreSQL-compatible context lake giving AI agents shared, live, transactional state at decision time
Reducto
Agentic document parsing API that turns complex PDFs, scans and spreadsheets into citation-grounded data for AI, now with the 1¢-per-page r-1 model
Euno
AI-native context platform that gives enterprise agents governed, trustworthy data
Prevalent AI
A sovereign enterprise knowledge graph that gives security teams and AI agents context