# LangParse

> LangParse is a document-intelligence API that turns PDFs, scans and images into clean, structured JSON with per-field confidence. It runs AWS-native, with data residency in the region you choose (US or Australia today). Use it to extract invoices, bank statements, contracts, forms and more — with a model you define, or schemaless (fields inferred automatically).

## How to use the API (for agents)

- **Base URL:** `https://api.langparse.dev`
- **Auth:** send an API key as `Authorization: Bearer <key>` or `X-Api-Key: <key>`. Keys carry scopes: `parse` (submit + read documents), `manage` (models/routers/sources/destinations), `admin`. New keys default to `parse`.
- **Response envelope:** every success response is wrapped — a single resource is `{ "data": { … } }`; a list is `{ "data": [ … ], "meta": { "pagination": { … } } }`. Errors are `{ "statusCode", "statusMessage" }`.
- **Parse a document** (one endpoint, async):
  1. `POST /api/v1/documents` with a `file` (`{ contents }` base64, `{ url }`, or `{ id }` from an upload) and **exactly one** strategy: `model: "mdl_…"`, `schemaless: true`, `auto: true`, or `router: "rtr_…"`. Optional `pages: [0,2]` (zero-indexed) to limit pages.
  2. It returns `202 { "data": { "id", "pollUrl" } }`.
  3. Poll `GET /api/v1/documents/{id}` until `status` is `processed`. The result carries `data` (the extracted values) and `fields` (per-field value + `confidence` + bounding boxes).
- **Large files:** `POST /api/v1/files` first (base64 / url / presigned), then parse by `file: { id }`.
- **Helpers:** `POST /api/v1/models/suggest` (design a schema from samples), `POST /api/v1/models/detect` (classify a file against your models).
- **Do not** combine strategies (e.g. `model` + `schemaless`) — that is a 400.

Example:

```bash
curl https://api.langparse.dev/api/v1/documents \
  -H "Authorization: Bearer $LANGPARSE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "file": { "url": "https://example.com/invoice.pdf" }, "schemaless": true }'
```

## Docs

- [Getting started](https://langparse.dev/docs/): what LangParse is and how to choose model vs schemaless.
- [Quickstart](https://langparse.dev/docs/quickstart/): submit a document and poll for the result in a few steps.
- [Core concepts](https://langparse.dev/docs/concepts/): models, async parsing, fields & confidence, routers, keys, the response envelope.
- [API Reference](https://langparse.dev/docs/api/): every endpoint, parameter and response, with a live console and multi-language code samples.

## Machine-readable

- [OpenAPI spec](https://api.langparse.dev/v1/openapi.json): the full OpenAPI 3.1 document.
- [Postman collection](https://langparse.dev/langparse.postman_collection.json): import to try every endpoint.

## Product

- [Solutions by industry](https://langparse.dev/solutions/): finance, healthcare, insurance, logistics, legal, real estate.
- [Use cases](https://langparse.dev/use-cases/): bank statements, PO↔invoice matching, variable layouts, multilingual, classification, human-in-the-loop.
- [Security](https://langparse.dev/security/) · [Enterprise](https://langparse.dev/enterprise/) · [Pricing](https://langparse.dev/#pricing)

## Get started

- [Sign up (free)](https://app.langparse.dev/signup)
- [Contact / sales](https://langparse.dev/contact-us)