Unstructured

What it is for

Unstructured is a document ETL desk: connect a source, partition messy PDFs into elements, enrich tables/images, chunk, and write to a vector store. Use it when the job is turn files into RAG-ready JSON at scale, not when you need a one-file Studio highlighter.

What makes it unique: the cover is the official Unstructured-IO/unstructured GitHub file list — unstructured/, example-docs/, Apache-2.0 in the About pane — not a marketing orb.

Key features

• Partition — split a file into titled elements (Auto / Fast / High-Res / VLM) (Pipelines walkthrough; pricing Partition Strategies). A parse, not a chat.
• Enrich — image description, table-to-HTML, generative OCR, NER on those elements (pricing Enrichment). A label pass, not a split.
• Chunk — by character, title, page, similarity, or contextual chunking (pricing Chunking Strategies). A slice, not a parse.
• Connectors — 20+ sources and destinations (S3, SharePoint, Weaviate, Elasticsearch…) so the job is a pipe, not an upload (pricing Source / Destination). A pipe, not a transform.
• Workflows — schedule, change-detect, and retry the same pipeline (pricing Workflow Management). A job, not a one-off file.

Embedding generation sits on the same transform. The Python SDK / Workflow API are install / SDK. Pages at $0.015 are the meter. VLM / model names are SKU.

Who it's for

Best fit: RAG teams who will burn the 10,000 free pages and then pay $0.015/page, not a team looking for a $20 chat seat.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from unstructured.io/pricing and the Apache-2.0 LICENSE.

• Software (OSS) — Apache 2.0. License fee $0. You host it.
• Let’s Go / start — 10,000 free pages when the hosted account starts. No card. All features.
• Pay-as-you-go — $0.015 / page after the first 10,000. All features.
• Business — custom. Dedicated instance / VPC / multi-tenant; multi-user; isolation.

Do not invent a monthly seat. Hosted dollars are per page. Open-source is free; you still pay compute.

Limitations / costs

• $0.015/page after 10k. There is no $20 Pro card.
• You bring a destination. Partition without Weaviate/S3 is only a demo.
• ETL, not a one-PDF Studio. For agentic parse boxes on a single file, see Reducto.

Alternatives

• Reducto — Studio parse/extract when you want a visual boxer, not 40 connectors.
• LlamaIndex — LlamaParse / indexes when Unstructured’s connector graph is not the job.
• Firecrawl — web crawl to markdown, not office/PDF ETL.

Official tutorial

Official Unstructured channel exists. Webinar: Unstructured 101 (59 min) is a talk, not a first-hour getting-started lesson — we do not embed it. Use written Pipelines walkthrough plus Pricing and that official channel.

Website

https://unstructured.io

Repository

https://github.com/Unstructured-IO/unstructured

Source

Catalog row curated for AI Tools Nav. Cover is the official Unstructured-IO/unstructured GitHub file list (Apache-2.0 + example-docs). Dollars checked on unstructured.io/pricing (10k free pages, then $0.015/page). Channel: @unstructuredio. No getting-started embed.

Last verified

2026-10-22 (Asia/Shanghai) editorial check.