Skip to main content
NDI (Nace Document Intelligence) is a platform API for document intelligence. Upload files to a workspace, perform granular document operations on any file, and query the corpus you build. The platform provides two integration patterns, plus published clients:

Granular operations

Parse, classify, extract, or ground individual documents without a workspace. Single operations, immediate results.

Workspace workflows

Upload files into persistent workspaces, trigger ingestion to build derived outputs, and use tools to search and analyze the corpus.

SDKs and MCP

Python and TypeScript clients, plus an MCP server for Claude Code, Cursor, and opencode.

​
Core capabilities

  • Parse — Convert any document into high-fidelity markdown. Spreadsheets become per-sheet tables. Audio and video are transcribed with speaker attribution.
  • Classify — Label a document or its pages against classes you provide.
  • Extract — Pull structured JSON from documents using a JSON Schema you supply.
  • Ground — Locate target text in documents. Each target returns a location status, ranked matches, and an exact, normalized, or semantic match method.
  • Search — Hybrid, semantic, and BM25 search over workspace files. Knowledge graph search for relationship queries.
  • Intelligent search — Agentic multi-step search that orchestrates tools to answer complex questions over a corpus.
  • File QA — Natural-language questions about any uploaded file — documents, spreadsheets, images, recordings — answered with citations.

​
Base URL

All endpoints are relative to this base. Authentication is a single header — see Authentication.

​
Async semantics

Every operation that takes more than a moment returns a Job resource. Each job-creating route accepts ?wait_seconds= (0–300) to optionally block for a result. See Jobs and idempotency.
Ready? Head to the Quickstart, or install the Python or TypeScript SDK.