Getting started

LangParse is a document-intelligence API. Send a document, get back clean, structured data — with per-field confidence, a review queue for anything low-confidence, and tools to measure accuracy over time.

What you can do

1. Parse documents and scans in almost any format — PDF, images, and more.
2. Define the fields you want back as a model (a document type + its schema), or extract schemaless and let LangParse infer the fields.
3. Extract reliably across variable layouts and 100+ languages, returning a consistent set of fields with confidence scores.
4. Route & split multi-document files, classify them, and send each part to the right model automatically.

Pick your path

• Just want structured data fast? → Quickstart — parse your first document in a few minutes.
• Want the mental model first? → Core concepts — models, async parsing, confidence, routers, and keys.
• Building an integration? → API Reference — every endpoint, parameter, and response, with a live “try it” console.
• Prefer Postman? → import the Postman collection or the OpenAPI spec.

Two ways to extract

Both go through the same endpoint — POST /api/v1/documents — you just choose how it parses:

	Model-based	Schemaless
Schema	You define the fields	Inferred automatically
Best for	Repeatable document types	One-offs, exploration, the live demo
Submit with	{ "file": …, "model": "mdl_…" }	{ "file": …, "schemaless": true }

Or pass "auto": true to detect the best model and parse in one step.

Everything is scoped to an organization with a data-residency region you choose (US or Australia today). Sign up free to get a key, then head to the Quickstart.