Llama Cloud Services is the open-source SDK repository for LlamaCloud, LlamaIndex’s managed platform for document intelligence. It contains the Python and TypeScript clients — and agent skills — for services such as LlamaParse, the agentic document parser that turns PDFs, Office files and presentations into clean, LLM-ready data. The project has about 4,300 stars on GitHub.
Originally the home of the LlamaParse client (the repository was renamed from llama_parse), the repo now covers the whole LlamaCloud family: LlamaParse for parsing complex documents with tables, charts and handwriting; LlamaExtract for pulling structured JSON against a schema; LlamaClassify and LlamaSplit for routing and segmenting document collections; and indexing/retrieval services for knowledge agents. The SDKs handle uploads, asynchronous job polling, webhooks and result downloads, and the included agent skills let coding agents and frameworks call the services directly. Typical pipelines feed enterprise document stores — contracts, filings, invoices, research reports — into RAG systems and extraction workflows where raw OCR text is not good enough. The libraries are MIT licensed and actively developed in 2026; parsing itself runs on LlamaIndex’s cloud.
The services are the commercial engine behind many production document-AI products: LlamaParse alone is widely used to feed retrieval systems in finance, legal tech and research, because it preserves the layout semantics — headings, tables, footnotes — that plain OCR discards. The SDKs expose the same jobs over REST, so anything you prototype with the client can be reproduced from other languages, and results can be delivered as markdown for chunking, JSON for databases, or raw text. Caching and page-level controls help manage cost on very large corpora.
The SDKs are free and open source (MIT). The LlamaCloud services they call are Freemium: new accounts receive free credits, with pay-as-you-go and enterprise plans beyond that.
Llama Cloud Services is for developers building document pipelines and RAG applications who want production-grade parsing of complex files without running their own OCR infrastructure.