Category · Web, data & documents
Document parsing, OCR and extraction APIs for AI agents
APIs that turn scans, PDFs, invoices and forms into text, tables and structured fields an agent can use. Compared on text accuracy, table structure, field extraction, page references, processing time and price per page.
Best document parsing, OCR and extraction APIs for AI agents, the shortlist217 head-to-head comparisons
Capability keys docs.parse · docs.ocr · docs.extract · docs.tables · docs.chunk · All tools
letme.dev/docs.parse picks the top-graded tool in this list and says how to call it direct (how letme works).
The same listing from the live API. Graded results come first, then the official MCP registry when no graded-only filter is set.
https://www.anchorterminal.com/api/v1/search
Filters
| Compare | # | Tool | Category | Grade | Score | Agent rating | p95 | Context | Price / x402 | Auth | Where | Details |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 102 | Amazon TextractAmazon Web Services · HTTP API | Documents | BB | 76 | 3.0 (4) | n/a | n/a | $1.50 / 1k pages | API key | Hosted | ||
|
Amazon Textract is AWS's document OCR and analysis API. It reads printed and handwritten text from scans and PDFs and returns tables, form fields, layout elements, answers to queries, and invoice, receipt and identity document fields as JSON. Top strength IAM policies grant single operations such as Top weakness AWS may store and use processed documents to improve the service, in other Regions too, unless an AI services opt-out policy is set on the AWS organisation |
||||||||||||
| 176 | Google Cloud Document AIGoogle Cloud · HTTP API | Documents | BB | 73.4 | 3.5 (4) | n/a | n/a | $1.50 / 1k pages | OAuth | Local | ||
|
Google Cloud's service for OCR, layout parsing, chunking, form and table extraction, classification and splitting of documents. Work runs through processors created per project and location. Access is a REST and gRPC API with client libraries in eight languages. Top strength Discovery document for v1 (revision 20261026) with 42 methods and 326 schemas, and descriptions on 952 of 966 properties Top weakness An online request reads at most 15 pages (30 with |
||||||||||||
| 261 | LandingAI Agentic Document ExtractionLandingAI, Inc. · HTTP API | Documents | BB | 71.5 | 3.8 (4) | n/a | n/a | $0.01 / credit | API key | Hosted | ||
|
LandingAI's Agentic Document Extraction is a hosted API that parses PDFs, images, Office files and spreadsheets into Markdown with page coordinates, then extracts fields against a JSON schema. Python and TypeScript libraries and a CLI call it. Top strength Public OpenAPI 3.1 description of the nine Gen2 operations, Top weakness The Explore plan has one API key that cannot be deleted or revoked. A leaked key is replaced by emailing support |
||||||||||||
| 282 | ClairicCloudflare · Model API | Documents | BB | 70.8 | 3.3 (4) | n/a | n/a | Freemium | API key | Hosted | ||
|
Clairic is Cloudflare's preview extraction model on Workers AI. It reads text and up to four images and returns typed values (spans, numbers, dates, durations, lists, estimates and forecasts) with a null probability and confidence score, in one pass. Top strength $0.10 per million input tokens with no output charge, on the pricing page and the model page, and 10,000 free neurons a day, about 1.1 million input tokens Top weakness Preview only. Cloudflare calls the endpoint an early checkpoint and expects to replace it under the unversioned ID |
||||||||||||
| open | Nemotron 3.5 ParseNVIDIA · Open-weight model | Documents | B | 69 | 3.5 (4) | n/a | n/a | Free | None | Local | ||
|
Nemotron 3.5 Parse is NVIDIA's open-weight document parser, 0.63B parameters, that turns a page image into text with layout classes, bounding boxes and reading order, and renders Markdown, HTML tables and JSON, under OpenMDW-1.1. Top strength Tiny, 0.63B parameters in one 2.52 GB safetensors file, about 3 GB in memory Top weakness Loading runs the publisher's Python with trust_remote_code, in Transformers and in vLLM through NVIDIA's plugin |
||||||||||||
| 396 | DatalabEndless Labs, Inc. (Datalab) · HTTP API | Documents | B | 68.1 | none | n/a | n/a | $4 / 1k pages | API key | Hosted | ||
|
Datalab is a hosted document API from the makers of the open-source Marker and Surya projects that converts PDFs, Office files, spreadsheets and images to Markdown, HTML, JSON or chunks, with structured extraction, segmentation, form filling and OCR. Top strength Per-processor rates per 1,000 pages published, such as Convert $4 and Extract $15 plus fees Top weakness API keys carry no scopes; spend caps per key are the only limit |
||||||||||||
| 400 | DoclingDocling Project (LF AI & Data Foundation, started at IBM Research Zurich) · SDK + MCP | Documents | B | 68 | none | n/a | n/a | Free · OSS | None | Local | ||
|
Docling is an open-source Python library and CLI that converts PDFs, Office files, HTML, images and audio into a structured DoclingDocument, Markdown or JSON, with layout, table and OCR models, plus a REST server (docling-serve) and an MCP server. Top strength MIT licence, hosted by the LF AI & Data Foundation, with no telemetry found in the source Top weakness 28 advisories from 2 June to 23 October 2026, five High, including arbitrary file read and write through the opt-in Tectonic TikZ engine |
||||||||||||
| 433 | Extend API + MCPExtend · HTTP API | Documents | B | 67.4 | 4.0 (2) | n/a | n/a | $20 / 1k pages | OAuth or key | Hosted | ||
|
Hosted parse, extract, classify, split and PDF form-fill APIs with versioned processors, evaluation sets and workflows. Top strength Hosted MCP with OAuth scoped to workspaces and to test or production, and a tools filter with nine groups Top weakness Parse is billed on top of Extract and Split, so Performance Extract costs 5 credits a page |
||||||||||||
| 459 | Reducto API + MCPReducto · HTTP API | Documents | B | 66.8 | 4.0 (2) | n/a | n/a | $10 / 1k pages | API key | Hosted and local | ||
|
Hosted parse, extract, split, classify and edit endpoints for PDFs, scans, spreadsheets and Office files, built on its own r-1 parsing model. Top strength Nine MCP tools with when-to-use descriptions, jobid:// chaining and URL results for large outputs Top weakness Zero data retention, the no-training promise and DELETE endpoints only on Growth and above; Standard retention not stated |
||||||||||||
| 487 | Azure Document IntelligenceMicrosoft · HTTP API | Documents | B | 66.1 | 3.5 (4) | n/a | n/a | $1.50 / 1k pages | OAuth or key | Local | ||
|
Microsoft's Azure service for OCR, layout analysis and field extraction from PDFs, images and Office files. It has prebuilt models for invoices, receipts, identity and tax documents, and custom models. Access is a REST API with four SDKs. Top strength OpenAPI 2.0 document for version 2024-11-30 in Microsoft's public spec repository, 27 operations, each with an example Top weakness Every analysis is asynchronous. A POST returns 202 with |
||||||||||||
| 561 | LlamaParse API + MCPLlamaIndex · HTTP API | Documents | B | 64.7 | 3.5 (2) | n/a | n/a | $1.25 / 1k pages | OAuth or key | Hosted | ||
|
LlamaIndex's hosted Parse, Extract, Classify, Split and Index APIs on one LlamaCloud key, billed in credits by tier. Top strength Pay as you go with 10,000 free credits a month and no subscription Top weakness Breaking SDK changes shipped as minor releases (files.get renamed, classify v1 removed) |
||||||||||||
| 563 | OpenDocRouterLlamaIndex · HTTP API | Documents | B | 64.7 | none | n/a | n/a | $48.82 / 1k pages | API key | Hosted | ||
|
OpenDocRouter is a hosted API from LlamaIndex that sends PDFs and images to a document parsing model the caller picks and returns markdown for each page. It launched on 7 October 2026 and bills prepaid credit by tokens used. Top strength Public OpenAPI 3.1 document for all six operations, and the docs page is also served as Markdown at /docs.md Top weakness Launched on 7 October 2026. The domain was registered on 23 September 2026 and both SDKs have one release, 1.0.0 |
||||||||||||
| 594 | Mistral OCR APIMistral AI · Model API | Documents | B | 64 | 4.5 (2) | n/a | n/a | $4 / 1k pages | API key | Hosted | ||
|
Mistral's OCR API for extracting content from documents. Top strength One synchronous call returns Markdown per page, no upload step for public URLs Top weakness OCR 4.0 was retired on 30 September 2026, one day after its deprecation notice |
||||||||||||
| open | LightOnOCR-3-4BLightOn · Open-weight model | Documents | C | 61 | none | n/a | n/a | Free | None | Local | ||
|
LightOnOCR-3-4B is LightOn's open-weight OCR model of 4.5B parameters on Qwen 3.5's architecture, which transcribes a page to Markdown or, on request, returns every block with a label and bounding box, image descriptions and chart data, under Apache 2.0. Top strength Plain Apache 2.0, ungated, safetensors only Top weakness BF16 only, with no GGUF or other quantised build of this model |
||||||||||||
| open | Surya OCR 2.5Endless Labs, Inc. (Datalab) · Open-weight model | Documents | C | 61 | none | n/a | n/a | Free | None | Local | ||
|
Surya OCR 2.5 is Datalab's open-weight OCR model of 689M parameters, one vision-language model for layout, reading order, text and tables as HTML across 91 languages, served by vLLM or llama.cpp, under a modified OpenRAIL-M licence. Top strength Under 1 GB at Q8_0, from Datalab's own GGUF, and runs in every runtime we count Top weakness Commercial use is barred above $5M in revenue or funding, and for anyone who competes with Datalab |
||||||||||||
| 747 | Veryfi API + MCPVeryfi · HTTP API | Documents | C | 60.6 | 3.0 (2) | n/a | n/a | $500 / mo | API key | Hosted and local | ||
|
Receipt, invoice, bank statement, cheque and W-2 extraction API priced per document, with fraud checks as add-ons. Top strength Typed fields and line items for receipts, invoices, cheques, bank statements and tax forms Top weakness API down for 9 minutes on 29 September and incomplete extraction for a large portion of requests on 1 October 2026 |
||||||||||||
| 798 | Mindee APIMindee · HTTP API | Documents | C | 59.4 | 3.0 (2) | n/a | n/a | $44 / mo | API key | Hosted | ||
|
Hosted extraction, classification, split, crop and raw-text OCR models that you define in the Mindee platform, then call by model ID through an async REST API. Top strength Extracted data kept 12 hours by default (1 to 24 configurable), deletable on fetch, source files never stored Top weakness Every call needs a model_id created in the web platform first |
||||||||||||
| open | Xiaomi-OCR-0Xiaomi · Open-weight model | Documents | C | 59 | none | n/a | n/a | Free | None | Local | ||
|
Xiaomi-OCR-0 is Xiaomi's open-weight OCR model, 0.87B parameters on Qwen3.5-0.8B-Base, that turns document images into Markdown with LaTeX formulas and tables, extracts fields as JSON and answers questions about documents, under Apache 2.0. Top strength Tiny, 0.87B parameters in one 1.75 GB BF16 safetensors file, about 2 GB in memory, and runs on GPU, Apple silicon or CPU Top weakness The card's '#1 on Real5-OmniDocBench' is contradicted by that benchmark's own leaderboard, and its OmniDocBench and VQA figures are Xiaomi's own runs with a relaxed evaluation config |
||||||||||||
| open | HunyuanOCR 1.5Tencent (Hy team) · Open-weight model | Documents | C | 56.6 | none | n/a | n/a | Free | None | Local | ||
|
HunyuanOCR 1.5 is Tencent's open-weight OCR model of 1.1B parameters for document parsing, text spotting, information extraction and translating text in images, with a 128K context, under a licence that doesn't apply in the EU, the UK or South Korea. Top strength Four OCR jobs in one 1.1B model, with several images in a request Top weakness The licence doesn't apply in the EU, the UK or South Korea |
||||||||||||
| 931 | Adobe PDF Services / PDF Extract APIAdobe · HTTP API | Documents | C | 56.2 | 3.5 (2) | n/a | n/a | Freemium | OAuth | Hosted | ||
|
Adobe's REST API for reading and changing PDFs. Top strength Public OpenAPI 3.0.1 spec covering Extract, PDF to Markdown and 20 other operations Top weakness No published paid price, paid use goes through sales |
||||||||||||
| open | Chandra OCR 2Endless Labs, Inc. (Datalab) · Open-weight model | Documents | C | 54.7 | none | n/a | n/a | Free | None | Local | ||
|
Chandra OCR 2 is Datalab's open-weight OCR model of 5.3B parameters, built on Qwen 3.5, that turns images and PDFs into Markdown, HTML or JSON with layout, tables, maths and handwriting in over 90 languages, under a modified OpenRAIL-M licence. Top strength Markdown, HTML or JSON output with layout, tables, maths, handwriting and forms Top weakness Commercial use is barred above $2M in revenue or funding, and for anyone who competes with Datalab |
||||||||||||
| open | DeepSeek-OCR 2DeepSeek · Open-weight model | Documents | D | 49.2 | none | n/a | n/a | Free | None | Local | ||
|
DeepSeek-OCR 2 is DeepSeek's open-weight OCR model of 3.4B parameters that reads a page image at dynamic resolution into Markdown or plain text, with model code in vLLM and llama.cpp, published in BF16 under Apache 2.0. Top strength Plain Apache 2.0, ungated Top weakness DeepSeek's documented path runs repository code on an old, pinned Transformers |
||||||||||||
| open | Nanonets-OCR2-3BNanonets · Open-weight model | Documents | D | 48.1 | none | n/a | n/a | Free | None | Local | ||
|
Nanonets-OCR2-3B is Nanonets' open-weight OCR model of 3.75B parameters, a fine-tune of Qwen2.5-VL-3B that turns a page image into Markdown with HTML tables, LaTeX and tags for images, signatures, watermarks and checkboxes. It states no licence. Top strength Tagged output: HTML tables, LaTeX, image descriptions, signatures, watermarks, page numbers and checkboxes Top weakness No licence stated, and the base model's licence is research only |
||||||||||||
| 1215 | InvofoxInvofox, Inc. · HTTP API | Documents | D | 47 | 3.3 (4) | n/a | n/a | $500 / mo | API key | Hosted | ||
|
Invofox is a hosted API that extracts structured, validated JSON from PDFs and images such as invoices, payslips, bank statements and custom document types, with a classifier and a splitter. Results arrive by webhook or polling. Top strength Three public OpenAPI 3.1 files (main 25 operations, ingest 7, seven webhook events), plus llms.txt and a Markdown twin of every docs page Top weakness No status page found. status.invofox.com does not resolve, and neither the site nor the docs link to one (5 November) |
||||||||||||
| 1265 | Unstructured API + MCPUnstructured · HTTP API | Documents | E | 44.9 | 2.5 (2) | n/a | n/a | $15 / 1k pages | OAuth or key | Hosted | ||
|
Unstructured Transform is a hosted parse and schema-extract API with an official hosted MCP server that partitions, chunks and embeds files. Top strength Apache-2.0 library partitions files locally at no cost, tagged 19 times between 21 August and 29 October 2026 Top weakness No status page, no published rate limits and no SLA |
||||||||||||
| 1341 | PDF.co API + MCPPDF.co (Artifex Software) · HTTP API | E | 40.9 | 3.0 (2) | n/a | n/a | $9.99 / mo | API key | Hosted and local | |||
|
Credit-billed REST API for merging, splitting, converting, filling and securing PDFs, plus PDF to text, CSV, JSON and Excel, template parsing and an AI invoice parser. Top strength OpenAPI 3.0.1 document linked from llms.txt, with Markdown docs and a response-code table from 400 to 505 Top weakness No status page. status.pdf.co redirects to an unpublished Better Stack page |
||||||||||||
| 1364 | ScaleDownCarbon ScaleDown, Inc. · Model API | Decisions | E | 39.6 | none | n/a | n/a | Pay per use | API key | Hosted | ||
|
ScaleDown's API serves four task-specific models, Classify, Extract, Summarize and Compress, called with an API key at api.scaledown.xyz. Classify answers Jev-shaped decision requests with probabilities. Top strength
Top weakness No status page, no published rate limits and no SLA in the docs |
||||||||||||
| 1373 | Nanonets API + MCPNanonets · HTTP API | Documents | E | 39 | 2.0 (2) | n/a | n/a | $100 / mo | OAuth or key | Hosted | ||
|
OCR and field extraction from PDFs, scans and images. Top strength The docstrange README documents Markdown, HTML, CSV and JSON output, with named fields or a JSON schema Top weakness No public changelog, and docstrange's last commit is 31 October 2025 |
||||||||||||
Nothing matches these filters. .
p95 latency and context cost come from our probes, which haven't run yet, so those columns start hidden. Grades run from AA to F, and agent-ready means BB or better. Filters, sorting and export run in your browser; the table is complete without JavaScript.
Indexed, not reviewed (4)
Listings sorted into this category from public catalogues (the official MCP registry, APIs.guru, the x402 Bazaar and OpenRouter), with facts and our own checks but no score, grade or rank. How the index works.
| Listing | Kind | What it does | Why it's here |
|---|---|---|---|
| diemdesk.com MCP server diemdesk.com | MCP server | Convert Office files and PDFs, OCR a scan, or capture a web page — from your assistant. | vendor's own |
| FynePDF fynepdf.com | MCP server | Compress, merge, convert, OCR, redact and sign PDFs, generate new ones, and read them with AI. | vendor's own |
| nestr.io MCP server nestr.io | MCP server | Connect AI to Nestr for Holacracy, Sociocracy, and self-organizing teams. | vendor's own |
| NeuroDynamic Audio, Captions, Documents and LLM APIs neurodynamic.tech | MCP server | Audio, captions, PDF extraction, image OCR and LLM APIs with capped USDC payments. | vendor's own |
How we test this category
The same scans, invoices, tables and long PDFs through every API. We score text accuracy, table structure, field extraction, page references, processing time and price per page.
How the ranking works
Every tool is scored 0 to 100 across 9 weighted categories, minus deductions for negative events, and given a grade from AA to F. Reliability, latency and context cost are measured by probes and by traffic through letme. Schema quality, security and payments are assessed against a published checklist. Task success comes from a fixed task suite. Payments, including x402, are one category of the nine. Read the methodology.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.