Document parsing APIs
A document parsing API takes a PDF, scan, slide deck or spreadsheet and gives back text an LLM can use: Markdown, JSON or chunks, with tables kept as tables and pages in reading order. Almost all of them bill per page, from $1.50 per 1,000 pages for plain OCR to $10 to $56 for layout, forms, agentic parsing and schema extraction. The cloud giants sell OCR plus form and layout models. LlamaParse, Unstructured and Reducto are built for RAG and agents. Mistral sells one OCR model at a flat price, and Docling is free and runs on your own machine. Run your ugliest documents through two or three before you commit: tables, handwriting and multi-column pages separate them far more than price does. 10 tools, in the order they appear in the directory. This is not a ranking.
Grouped by job: AI Agent Infrastructure
Invofox
Document parsing API that turns invoices, payslips, contracts and any document into validated JSON or Markdown for RAG and LLM pipelines, with classification, splitting, confidence scores and continuous learning; SOC 2 and ISO 27001.
Firecrawl
Web data API for AI agents: scrape any URL into clean Markdown or JSON, crawl and map whole sites, search the web, run browser sessions and monitor pages, with an MCP server. Document parsing too: PDFs, including scans with OCR, and Word, Excel and PowerPoint files to Markdown at 1 credit a PDF page. Open source under AGPL-3.0 with a self-host guide, or use the cloud with 1,000 free credits a month.
LlamaParse
LlamaIndex's document parsing API for 130+ file types: PDFs, scans, slides and spreadsheets come back as Markdown, text, JSON or HTML tables, in four tiers from Fast to Agentic Plus. Billed in credits at $1.25 per 1,000, from 1 to 45 credits a page, with 10,000 free credits a month.
Unstructured
Document parsing for RAG: partition 50+ file types with Fast, High-Res or VLM strategies, then chunk, enrich and embed, with 40+ source and destination connectors. Hosted at $0.015 a page after 10,000 free pages, or the open-source library under Apache 2.0.
Reducto
Document parsing API for PDFs, scans, spreadsheets and 30+ file types: Parse captures layouts and tables with bounding-box citations, Extract fills your JSON schema, plus Split, Classify and Edit. Parse is $10 per 1,000 pages, with $150 of free usage and VPC or on-prem on Enterprise.
Mistral OCR
Mistral's OCR model for document parsing: OCR 4.1 reads PDFs, PPTX, DOCX and images into Markdown page by page, with tables as Markdown or HTML, headers and footers, bounding boxes and confidence scores, in 40+ languages. $4 per 1,000 pages, $5 with structured annotations.
Google Document AI
Google Cloud's document parsing service: Enterprise Document OCR at $1.50 per 1,000 pages, Layout Parser with chunking for RAG at $10, Form Parser and custom extractors at $30, plus prebuilt invoice and expense parsers. No charge for failed requests.
Azure AI Document Intelligence
Microsoft's document parsing service, now part of Foundry Tools: Read OCR at $1.50 per 1,000 pages, a Layout model that outputs Markdown with tables at $10, and prebuilt and custom extraction models. 500 free pages a month, and containers to run it on your own hardware.
Amazon Textract
AWS's document parsing and OCR service: Detect Document Text at $1.50 per 1,000 pages, Analyze Document for tables, forms, layout and natural-language queries, plus Expense and ID APIs. New AWS customers get a three-month free tier.
Docling
Open-source (MIT) document parsing library started at IBM Research: converts PDF, DOCX, PPTX, XLSX, HTML and images to Markdown, HTML or JSON with layout, reading order, tables, formulas and OCR for scans. Runs locally, with docling-serve for an HTTP API and an MCP server.
What it is, what comes out, what 1,000 pages cost.
Straight from each vendor's own pages. Plans move often, so the check date is below.
| Tool | What it is | Output | Price per 1,000 pages | Free to start |
|---|---|---|---|---|
| Invofox | Extraction API for invoices, receipts, payslips, contracts and other business documents. | Validated JSON or Markdown, with confidence scores. | Starter from $500 a month billed annually, up to 5,000 pages a month. Scale and Enterprise quoted. | 500 pages, no card. |
| Firecrawl | Web data API that also parses uploaded files and document URLs. | Markdown per page, plus layout blocks with bounding boxes. | 1 credit per PDF page. Hobby $19 a month for 5,000 credits, Standard $99 for 100,000. | 1,000 credits a month. |
| LlamaParse | LlamaIndex's hosted parser for 130+ file types, in four tiers from Fast to Agentic Plus. | Markdown, text, JSON, XLSX and HTML tables. | Credits at $1.25 per 1,000. Fast $1.25, Cost-effective $3.75, Agentic $12.50, Agentic Plus $56.25. Starter $50 a month for 40,000 credits. | 10,000 credits a month. |
| Unstructured | Hosted partitioning, chunking, enrichment and embedding for 50+ file types, with 40+ connectors. | JSON elements, chunked and optionally embedded. | $15 ($0.015 a page). Business plans with VPC on request. | 10,000 pages, no card. The open-source library is free under Apache 2.0. |
| Reducto | Parse, Extract, Split, Classify and Edit APIs on its r-1 models, for 30+ file types. | Structured output with bounding boxes. Extract fills your JSON schema. | Parse $10, Extract $20 with parsing included, Deep Extract $40, Classify $7.50. Batch takes 20% off Parse and Extract. | $150 of usage on Standard. |
| Mistral OCR | Mistral's OCR 4.1 model on the Mistral API, also in batch. | Markdown per page, tables as Markdown or HTML, bounding boxes, confidence scores. | $4, or $5 with annotations. | Not listed. |
| Google Document AI | Google Cloud processors for OCR, layout parsing with chunking, forms, and prebuilt and custom extractors. | Document JSON with text, layout and fields. Layout Parser adds chunks. | Enterprise OCR $1.50 ($0.60 above 5M pages), Layout Parser $10, Form Parser $30, custom extractor $30. | First 1,000 OCR pages in the price table. Failed requests are free. |
| Azure AI Document Intelligence | Read, Layout, prebuilt and custom models in Microsoft Foundry Tools, also as containers. | JSON, and Markdown with tables from Layout. | Read $1.50 ($0.60 after 1M pages), Layout and prebuilt $10, custom extraction $30, add-ons $6. East US list prices. | 500 pages a month on the free F0 tier. |
| Amazon Textract | AWS APIs for text, tables, forms, layout and queries, plus receipts and IDs. | JSON blocks with page geometry. | Detect Text $1.50 ($0.60 after 1M), Tables $15, Forms $50, Queries $15, Layout free with Tables. US West (Oregon). | New customers, 3 months: 1,000 pages a month of Detect Text, 100 of Forms, Tables or Layout. |
| Docling | Open-source Python library started at IBM Research, now an LF AI & Data project. | Markdown, HTML, JSON or DocTags, with a hybrid chunker. | Free. You pay for the machine it runs on. | Free under MIT. Self-host, or run docling-serve for an HTTP API. |
Checked 2026-11-02. Sources: LlamaIndex pricing, LlamaParse pricing docs, Unstructured pricing, Unstructured on GitHub, Reducto pricing, Mistral OCR docs, Mistral OCR 4.1 model, Google Document AI pricing, Azure Document Intelligence pricing, Azure Layout model docs, Amazon Textract pricing, Docling on GitHub, Firecrawl document parsing docs, Firecrawl pricing, Invofox pricing.
Pick it if.
- LlamaParseYou build on LlamaIndex, or want one parser with a cheap tier for clean files and an agentic tier for hard ones.
- UnstructuredYou want parsing, chunking and embedding in one pipeline, with connectors at both ends, or the open-source library to start.
- ReductoAccuracy on messy tables and forms matters more than price, and you want schema extraction on the same API.
- Mistral OCRYou want one flat price, $4 per 1,000 pages, and Markdown out with no tiers to pick.
- Google Document AIYou're on Google Cloud and want OCR, layout chunks and form parsing in one account.
- Azure AI Document IntelligenceYou're on Azure, want Markdown from the Layout model, or need to run it in your own containers.
- Amazon TextractYour documents already sit in S3 and you want forms, tables and queries on your AWS bill.
- DoclingDocuments can't leave your machines, or you'd rather pay for compute than per page.
- FirecrawlYou already scrape the web with Firecrawl and want PDFs and Office files on the same credits.
- InvofoxYou process invoices, receipts or payslips and want validated fields, not just text.
Each listing.
The directory page, alternatives in the same category, and a head-to-head where we have one.
- InvofoxAlternatives
- FirecrawlAlternatives
- LlamaParseAlternatives
- UnstructuredAlternatives
- ReductoAlternatives
- Mistral OCRAlternatives
- Google Document AIAlternatives
- Azure AI Document IntelligenceAlternatives
- Amazon TextractAlternatives
- DoclingAlternatives
Category pages list every tool in that category, including ones left off this page. Built one of these? Get it featured in its category.
The 2 biggest AI stories, every morning.
One short email: the two stories that matter, and why.
You're in.
The Daily AI Brief will land in your inbox soon. If you don't see it, check Promotions or Spam and move it to your main inbox.