> ## Documentation Index
> Fetch the complete documentation index at: https://docs.abbyy.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

FineParser is a self-hosted document parser for AI pipelines. It runs as one Docker container on CPU hardware and exposes a small REST API: submit a document, then download its structure and text as DocLang, JSON, or plain text. No cloud account, no GPU, and no document data leaves your infrastructure.

It is built on the [ABBYY FineReader Engine](/fine-reader/engine/introduction) recognition core, packaged for developers who want production-grade parsing without an SDK integration.

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/fine-parser/getting-started/quickstart">
    Run the container and parse your first document in under five minutes.
  </Card>

  <Card title="Configuration" icon="sliders" href="/fine-parser/basics/configuration">
    Recognition mode, languages, and ports.
  </Card>

  <Card title="REST API" icon="code" href="/fine-parser/concepts/rest-api">
    Submit a document, poll the job, download the result.
  </Card>

  <Card title="Output formats" icon="file-code" href="/fine-parser/basics/output-formats">
    DocLang, JSON, and plain text.
  </Card>
</CardGroup>

## What it does

FineParser recovers a document's structure, not just its text: tables (including borderless ones), multi-column layouts, headings, reading order, and hierarchy. Handwriting, checkboxes, and images are recognized alongside the text. The output is ready for RAG ingestion, agent workflows, or LLM prompts.

## Why teams use it

* Keeps tables, columns, and headings intact, and tells you where each one is on the page.
* Outputs [DocLang](/fine-parser/concepts/doclang), a compact format built for LLM input and the one we recommend. JSON and plain text are also available.
* Runs as a single container on CPU hardware, on your own machines or in your own cloud.
* Two [recognition modes](/fine-parser/basics/recognition-modes): fast mode for high volumes, high-fidelity mode when accuracy matters most.
* Recognizes more than 200 [languages](/fine-parser/basics/languages), including Latin, Cyrillic, CJK, and Arabic scripts.
* Start on the free tier, upgrade to a paid plan as your volume grows, or move to [Enterprise](/fine-parser/reference/enterprise) for high throughput and offline deployment.