Nutrient DWS Data Extraction API

What it does

The Data Extraction API takes PDFs, scanned PDFs, images and Office files and returns either Markdown or spatial JSON that includes elements, bounds, confidence scores and page context. You choose a processing mode per request: text for born-digital PDFs and Office files, structure for simpler scans, understand for forms and complex layouts, or agentic for cursive handwriting and degraded scans. Auto mode routes each page to one of the three depth modes. An extract request can pull named fields according to a schema.

Developers call the parse endpoint with an API key and a file, and receive JSON with the Markdown in an output field. Documented examples cover curl, Java, C#, JavaScript, Python and PHP. Nutrient also offers an MCP server that lets Claude, Claude Code or Codex extract fields with citations, and a macOS desktop app, Extract Studio, that runs the engine on Apple silicon Macs. Industry pages cover invoices, mortgages, medical claims and similar document types.

The API suits developers and teams building document pipelines, such as retrieval-augmented search or structured data capture. Access requires a Nutrient account and an API key; free signup is available, but the pricing page conflicts on included credits (5,000 in the FAQ versus 0 in plan displays), so confirm the account allowance. Credits are billed per page by mode, so costs vary with the mode chosen. Documents are processed in Frankfurt, but AI inference for extract and agentic parsing may run in the US, and the free plan enables internal retention of inputs for product improvement by default.

What it can do

• Parse PDFs, scans, images and Office files into Markdown or spatial JSON
• Four processing modes: text, structure, understand and agentic, plus auto routing
• Extract named fields according to a schema, with confidence scores and source coordinates
• Split merged documents into logical sections and classify documents into categories
• Detect form fields, apply AI labeling and fill form values

Before you choose

• Mode choice drives cost: text is 1 credit per page, understand is 9 and agentic is 18, and extract adds 6 credits per page.
• Free-plan data retention for internal product improvement and model training is enabled by default; paid plans disable it by default.
• AI inference for extract and agentic parsing may run in the US, and no customer-selectable EU-only option is stated.

Listed under

• Data
• Data extraction

Get started

Google or GitHub signup is advertised for free monthly credits; email signup has watermarks and a 5 MB limit. Confirm the allowance because the pricing page lists both 5,000 and 0 credits.

1. Create an account and get an API key. The signup page directs users to Google or GitHub for free monthly credits and higher file limits; email signup has watermarks and a 5 MB file limit. Confirm the actual credit allowance because the pricing page conflicts.
2. Send a file to https://api.nutrient.io/extraction/parse with an Authorization Bearer header and an instructions field that sets the mode and output format.
3. Optionally connect the Nutrient MCP server to Claude or Codex with the documented one-URL setup and sign in with your Nutrient account.

Credits are billed per page by mode. Pay-as-you-go overage is billed up to a spending cap, and free credits are used before wallet funds. Unused credits do not roll over. Annual billing is paid upfront at a discount.

Source: nutrient.io/api/pricing/data-extraction-api, checked 1 Nov 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 1 Nov 2026. Each source supports part of the overview; availability and prices can change.

• Data extraction API for PDFs, scans, and Office | Nutrient ↗www.nutrient.io
• Data Extraction API pricing — credit plans | Nutrient ↗www.nutrient.io
• Sign up - Nutrient Document Web Services ↗dashboard.nutrient.io

Company website

nutrient.io: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 1 Nov 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.