LlamaParse, the hosted document platform, plus two open-source projects: LiteParse (local parsing) and the LlamaIndex Framework (agents and RAG). Find out which one you need.
Documentation for LlamaParse, the hosted platform for parsing and extracting complex documents, and for the two open-source projects beside it: LiteParse and the LlamaIndex Framework.
Hosted platform
Agentic OCR and document processing as a managed API: Parse, Extract, Classify, Split and Index, with SDKs for Python, TypeScript, Go and Java.
Best for Production pipelines and the hardest documents: scans, forms, tables, charts, 130+ formats.
Open source · runs locally
A fast Rust parser that runs on your machine: layout, bounding boxes, Markdown, built-in OCR. No cloud, no API key.
LiteParse docs →
Open source · Python & TypeScript
Build agents and RAG over your data: workflows, retrieval, hundreds of integrations. Works with either parser.
Framework docs →The two parsers turn files into text and structure. The framework turns that into an application. Most production stacks use the framework with one of the parsers; many use LlamaParse on its own.
| I want to… | Start with |
|---|---|
| Parse PDFs, scans and forms with the best possible accuracy | LlamaParse |
| Turn documents into structured JSON that matches my schema | LlamaParse · Extract |
| Run document processing inside my own cloud | LlamaParse · Self-hosting |
| Parse documents locally, offline, or inside a coding agent | LiteParse |
| Get text with bounding boxes, fast, without any API calls | LiteParse |
| Build an agent or a RAG application in Python or TypeScript | LlamaIndex Framework |
