Nutrient Data Extraction API: JSON Matching Your Schema With Per-Field Citations and Confidence
Summary
- Nutrient has launched a Data Extraction API that parses PDFs, scans, images, and Office files into spatial JSON or Markdown matching a specified schema.
- The API provides per-field citations including page numbers, bounding boxes, match labels, and confidence scores.
- It offers multiple parse modes (text, structure, understand, agentic) alongside classify and split endpoints for document workflows.
- Nutrient's Agentic setting scored highest on its WholeDoc Bench evaluation at 86.2 overall.
- The system supports integration via hosted APIs, an MCP server, agent skills, an Apple Silicon Mac app, and a self-hosted SDK.
Why it matters
Accurate document extraction with per-field grounding and confidence scores directly addresses costly data hallucination errors in enterprise AI pipelines, enabling reliable automation for complex document processing.