# LlamaIndex LlamaIndex is an American AI company based in San Francisco that develops agentic OCR and document intelligence infrastructure, plus an open-source data framework for connecting large language models to private data sources. Founded in 2023, it offers tools for data ingestion, indexing, and retrieval in LLM applications. LlamaIndex is an American artificial intelligence company headquartered in San Francisco, California. The company specializes in agentic optical character recognition (OCR) and document intelligence infrastructure, developing tools and frameworks that enable [large language models](https://www.wikiprompt.org/wiki/large-language-model) (LLMs) to ingest, index, and query structured and unstructured data sources. Its products are used to build applications that connect generative AI models to proprietary or private data, addressing limitations in model context windows and data accessibility. The company's flagship open-source framework, also named LlamaIndex, provides a Python-based toolkit for constructing retrieval-augmented generation (RAG) applications and AI agents. It includes data connectors, index structures, query engines, and agent orchestration primitives. In addition, LlamaIndex offers commercial products such as LlamaCloud, a managed service, and LlamaParse, an agentic OCR system that uses a multi-agent pipeline to process complex document layouts. ## History LlamaIndex traces its origins to a side project begun in late 2022 by Jerry Liu, then a machine learning engineering manager at AI safety startup Robust Intelligence and formerly a research scientist at Uber's autonomous vehicle division. While experimenting with [OpenAI](https://www.wikiprompt.org/wiki/openai)'s GPT-3 model, Liu encountered its inability to work reliably with private or proprietary data and its narrow 4,096-token context window. To address this, he built a small indexing utility and published it to GitHub in November 2022 under the name GPT Tree Index. Within a few months, the project had accumulated over 16,000 GitHub stars, 200,000 monthly downloads, and a Discord server of roughly 6,000 developers, which convinced Liu to turn it into a company. Simon Suo, a former Uber colleague who later worked at autonomous-driving startup Waabi, joined as co-founder. The company was formally incorporated in April 2023, and the project was rebranded LlamaIndex. In June 2023, the company announced an $8.5 million seed round led by Greylock Partners, with participation from Jack Altman, Lenny Rachitsky, and Charles Xie. The following year, a survey published by InfoWorld described LlamaIndex as an orchestration provider that helps connect LLMs to private data sources, noting its primary focus on data ingestion, indexing, and retrieval. In March 2025, LlamaIndex, together with Cisco, LangChain, Glean, and Galileo, participated in the launch of AGNTCY, an open-source initiative intended to support interoperability and collaboration among AI agents. In the same month, LlamaIndex announced a $19 million Series A funding round, bringing its total disclosed funding to approximately $27.5 million. Investors have included Databricks and KPMG, which made a minority equity investment as part of a commercial partnership. LlamaIndex simultaneously launched LlamaCloud, a managed cloud service based on the LlamaIndex framework. The company also published LlamaParse, an agentic OCR system that uses a multi-agent pipeline combining computer vision, specialized vision-language models, and LLM-based reasoning to handle document layouts, tables, charts, and handwriting that may be difficult for conventional OCR systems to handle. In October 2025, IBM released an open-source Python connector enabling IBM Db2 to serve as a vector store within LlamaIndex workflows. In March 2026, LlamaIndex released LiteParse, an open-source, TypeScript-native parsing library designed for local, offline execution. LiteParse runs entirely on the user's machine with no external API dependencies or Python requirements. Later, LiteParse was re-launched as a Rust project with native TypeScript and Python bindings, in addition to adding markdown output. In April 2026, LlamaIndex introduced ParseBench, a benchmark for evaluating document-parsing systems on enterprise documents. ## Open-source framework The original open-source LlamaIndex framework was developed as a Python library, in addition to its OCR products, for building retrieval-augmented generation (RAG) applications and AI agents. The framework provides data connectors, index structures, query engines, and agent orchestration primitives. LlamaIndex introduced the Workflows abstraction, an event-driven system for creating multi-step agent pipelines with durable state, enabling long-running document processing tasks to persist across sessions. The open-source libraries are commercially licensed without restriction, allowing broad adoption in enterprise and research settings. The framework is designed to work with various [AI](https://www.wikiprompt.org/wiki/artificial-intelligence) models and data sources, including [Amazon Web Services](https://www.wikiprompt.org/wiki/amazon-web-services), [Azure](https://www.wikiprompt.org/wiki/azure), and [Google Cloud](https://www.wikiprompt.org/wiki/google-cloud) storage services, as well as vector databases and traditional databases. It supports integration with [OpenAI](https://www.wikiprompt.org/wiki/openai) models and other LLM providers, enabling developers to build applications that query private documents, databases, and APIs. ## Products and services LlamaCloud is a managed cloud service that provides a hosted version of the LlamaIndex framework, offering scalable data ingestion, indexing, and retrieval capabilities. It is designed for organizations that require production-grade infrastructure without managing their own deployment. LlamaParse, the agentic OCR product, uses a multi-agent pipeline to extract text and structure from complex documents, including those with tables, charts, and handwriting. LiteParse, released in 2026, is an open-source parsing library that runs locally and offline, with a Rust core and bindings for TypeScript and Python. ParseBench, introduced in April 2026, is a benchmark for evaluating document-parsing systems on enterprise documents. It provides a standardized set of tasks and metrics to compare the performance of different OCR and parsing tools. ## Publications and research LlamaIndex has been the subject of several scientific research papers, including studies on extended context for InstructGPT, document-grounded response generation in information-seeking dialogues, and contextual querying in additive manufacturing. A 2025 paper from the Korean Institute of Smart Media described a patient record-combined emergency rescue guide built upon LlamaIndex, while another 2025 study evaluated large language models for information extraction from unstructured electronic health records in residential aged care. These publications demonstrate the framework's use in domains such as healthcare, manufacturing, and conversational AI. The company's tools have been adopted in academic and industrial research settings, particularly for applications requiring integration of LLMs with domain-specific data sources. The framework's open-source nature has facilitated its use in a wide range of projects, from simple document Q&A systems to complex multi-agent workflows. ## See also - [Retrieval-augmented generation](https://www.wikiprompt.org/wiki/retrieval-augmented-generation) - [AI agent](https://www.wikiprompt.org/wiki/ai-agent) - Data ingestion - Document intelligence --- Source: https://www.wikiprompt.org/wiki/llamaindex License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/) Last updated: 2026-09-05T13:22:18.626851+00:00