LlamaParse is a document parser made by LlamaIndex. Parsing here means reading a file such as a PDF, an Office document, a spreadsheet or a scanned image, detecting the structure of each page (headings, columns, tables, charts) and writing the content out in a form that a large language model or another program can use.
A program sends a document through the REST API, or a person uploads it in the web interface, and LlamaParse returns Markdown, plain text, JSON per page, XLSX or HTML tables. It handles headers, footers and split sections, extracts tables and charts, and reads handwriting and checkboxes. Parse tiers (Cost Effective, Agentic, Agentic Plus) trade cost against accuracy: the agentic tiers use large language models or vision-language models on each page. The same platform offers Extract (structured fields from documents), Classify, Split and Index for retrieval.
LlamaParse is a commercial hosted service, not open source, with credit-based pricing: each parsed page costs credits, and monthly plans include a number of credits. Enterprises can also deploy it in a private cloud.