LlamaIndex

LlamaIndex is two products under one name.

The first is the open-source framework that made the project well known: packages for Python and TypeScript that connect language models to your own data, build retrieval pipelines and run agentic workflows.

The second is LlamaParse, the hosted commercial product the company now leads with, described on its own site as agentic document processing and positioned as the piece that makes the framework usable on real business documents - product PDFs, contracts, invoices, scanned forms, decks and support articles.

The homepage frames the promise as turning any document into AI-ready context: layout-aware OCR, schema-based extraction and an index pipeline built for retrieval, with task-specific agents splitting text, tables and charts between specialists and auto-correction loops catching errors on messy scans. For marketers the value is concrete rather than abstract.

Most teams that try to build a chatbot or an assistant over their own content stall at the same place: their knowledge lives in PDFs, slide decks and spreadsheets that a model cannot read cleanly.

LlamaParse is aimed exactly there - Parse handles 50+ unstructured file types including embedded images, multi-page tables and handwriting; Extract turns documents into structured fields from a schema with no model training; Split and Classify segment and route documents using plain-language rules; and Index runs chunking and embedding for retrieval.

The published use cases are financial due diligence, invoice processing, technical document search and customer support, and the industry pages cover finance, insurance, manufacturing and healthcare.

A useful bonus for teams that want to keep data in-house is LiteParse: open-source local parsing, installable from npm, that handles PDFs, Office documents and images without sending anything to the cloud or spending LLM tokens, and returns bounding boxes so you can show sources. Pricing was readable on llamaindex.ai/pricing on 23 September 2026.

Free is $0 a month with 10,000 credits, which the homepage equates to roughly 1,000 pages, including agentic OCR, schema-based extraction and the ability to build end-to-end document agents; Starter is $50 a month with 40,000 credits and pay-as-you-go up to 400,000 credits; Pro is $500 a month with 400,000 credits and pay-as-you-go up to $5,000 a month, plus priority Slack support, with a limited-time bonus of 800,000 credits for business email domains; Enterprise is custom with volume discounts, five times higher rate limits, SSO, SaaS or hybrid cloud deployment and a dedicated account manager.

All self-serve tiers list a 100-user cap, credit pricing is published as 1,000 credits for $1.25, and there is a startup programme for free credits.

The company also publishes scale numbers on its homepage - 1 billion documents processed, 25 million package downloads a month and 300,000 LlamaParse users - alongside the claim that the parser is four times cheaper than frontier labs and five times more accurate than other APIs, which is vendor positioning rather than an independent benchmark.

Two cautions shape the fit.

First, this is a developer product: the platform UI handles parsing, extraction and indexing, but anything resembling a marketing workflow is assembled in code, so budget engineering time or stay with the no-code agent platforms already in this directory.

Second, credit consumption tracks document complexity, and a pipeline that re-parses large scanned archives every month will spend real money - the free tier is a genuine evaluation grant at roughly 1,000 pages, not a production allowance.