LlamaIndex

Turn complex documents into reliable, structured content that AI agents and applications can use.

Parse documents locally and quickly into AI-ready text and structured output, without cloud services, API keys, or LLM tokens.

Build context-aware AI applications and agents faster with reusable components, flexible integrations, and support for retrieval-augmented generation.

Build controlled, stateful AI workflows that coordinate agents, human review, and application logic across multi-step processes.

Build and deploy document agents that automate reliable workflows from document ingestion and extraction through reasoning and action.

LlamaParse

LlamaParse is an agentic document processing platform that turns complex documents into AI-ready content and structured data. Parse converts documents into formats such as Markdown, HTML tables, and JSON while handling layouts, tables, charts, images, handwriting, and other content. Extract returns fields and records according to user-defined schemas, with citations and confidence scores. The platform also includes tools to classify documents, split files into logical sections, and index content for search and retrieval. Developers can use its APIs, SDKs, web app, CLI, and MCP server. Deployment options include managed and single-tenant SaaS, customer cloud environments, and on-premises hosting. OCRia-1, a vision-language model powering LlamaParse, is designed for accurate parsing of complex documents, including tables and charts, and can be deployed in a customer's environment.

• Convert Complex Documents Into AI-Ready Content
  Transform complex or unstructured files into content that downstream AI applications and agents can use, while retaining important document structure and information.
• Extract Structured Data With Custom Schemas
  Specify the fields and records your workflow needs, then receive structured JSON that can be used in downstream applications and processes.
• Parse Complex Documents Accurately
  OCRia-1's reported benchmark results highlight its ability to parse varied document content, including text, formatting, tables, and charts.
• Preserve Complex Table Relationships
  Retaining table structure and connections between values and labels helps make complex tables interpretable for downstream document workflows.
• Trace Extracted Fields To Their Sources
  Use citations and confidence scores to inspect the source and certainty of extracted fields, supporting verification and human review.
• Balance Processing Cost And Accuracy
  Choose parsing settings to suit document requirements and budgets, or use automatic page-level tier routing. OCRia-1 is presented as an economical parsing option, with a stated cost of less than half a cent per page.
• Prepare Document Content For Search And Retrieval
  Organize parsed content with chunking and embeddings so it can be used in search, retrieval, and RAG workflows.
• Integrate Document Processing Into Existing Workflows
  Access Parse and Extract through developer tools and agent integrations, and connect document processing to existing applications and workflows.
• Deploy In The Environment That Fits Your Needs
  Choose a managed service or deploy in a dedicated, customer-controlled cloud or on-premises environment. OCRia-1 is available through LlamaParse or for deployment in a customer's environment.
• Process Documents Across 130+ File Formats
  Bring varied document and media formats into a common processing workflow, including scanned documents and files containing handwriting.
• Detect And Correct Parsing Errors Automatically
  Use recursive checks to identify and correct parsing errors, supporting processing of challenging scans and multimodal documents.
• Turn Charts Into Structured Data
  Convert chart and graph content into structured data for use in downstream document workflows.
• Categorize Documents Using Natural-Language Rules
  Apply natural-language rules to categorize documents as part of document processing and workflow automation.
• Separate Combined Files Into Logical Documents
  Turn combined files into individual documents or meaningful sections to prepare them for subsequent processing.
• Process Documents In 100+ Languages
  Support document processing across a broad range of languages with out-of-the-box multilingual compatibility.
• Process Documents With Low Latency
  OCRia-1 is positioned for low-latency document parsing, supporting faster processing of document workloads.

LiteParse

LiteParse is an open-source document parser from the team behind LlamaParse. It processes PDFs, Office documents, and images locally, without cloud services or API calls, and produces text and structured output for AI agents and programmatic workflows. Its local parsing combines embedded-text extraction with OCR for scanned content, and provides positional data such as bounding boxes. LiteParse is available through a CLI and developer libraries, with browser-based parsing also offered through a WebAssembly demo.

• Keep Document Processing On Your Machine
  Parse documents locally without a cloud dependency or API calls, keeping document data on the machine.
• Parse Documents With One Tool
  Handle PDFs, Office documents, and images through one interface and a consistent parsing pipeline.
• Extract Text From Scanned Documents
  Combine embedded-text parsing with OCR for scanned content. Use built-in Tesseract.js or connect an external OCR server.
• Locate Text In Source Documents
  Use bounding boxes and page positions to identify where parsed text appears in its source document, supporting citations and downstream workflows.
• Send Parsed Content To Downstream Tools
  Choose Markdown, plain text, or JSON output and use parsed content in programmatic workflows, including pipelines that send it to LLMs or vector stores.
• Integrate Parsing Into Developer Workflows
  Run LiteParse from the CLI or use its developer libraries and Python wrapper in scripts, CI/CD pipelines, agent workflows, and batch processing.
• Use An Open-Source Parser
  Access the open-source parser and its repository on GitHub.
• Capture Pages For Visual Analysis
  Pair extracted text with page screenshots when an agent needs visual context for deeper multimodal reasoning.
• Parse PDFs Directly In Your Browser
  Use the browser demo to parse PDF bytes locally with WebAssembly, without requiring a server.

LlamaIndex framework

LlamaIndex is an open-source framework for building context-aware AI applications and agents. It provides extensible building blocks such as memory, state management, and human-in-the-loop capabilities, plus Python and TypeScript SDKs and integrations with LLMs, tools, and data sources. Developers can use it to create retrieval-augmented generation pipelines and production-ready AI applications, with control to customize components and workflows.

• Customize Agent Components
  Use and adapt core agent components such as memory, state management, and human-in-the-loop capabilities to fit specific application workflows.
• Connect Models, Tools, And Data
  Build applications with connections to LLMs, tools, and data sources, including support for new releases and integrations.
• Build Context-Aware RAG Applications
  Create and customize retrieval-augmented generation pipelines that bring data into AI applications as relevant context.
• Orchestrate Multi-Step Agent Workflows
  Coordinate steps, loops, parallel paths, and multiple agents with control over state transitions and execution flow.
• Build With Python Or TypeScript
  Use supported Python and TypeScript SDKs to develop applications and integrate the framework into an existing development stack.
• Build And Deploy Without Restrictions
  Use the open-source framework to build applications with control over implementation and without restrictions on production or commercial use.
• Develop And Debug RAG Pipelines
  Use framework resources and components to create, trace, and debug retrieval-augmented generation applications.

Workflows

Workflows is an open-source framework for building controlled, multi-step agentic systems. Developers can chain steps, add loops and parallel paths, and combine multiple agents in workflows that can start, pause, and resume statefully. Workflows also supports asynchronous execution, type-safe control over data passed between steps, and human-in-the-loop review. It can be used with Python and TypeScript applications.

Category: AI Agent Orchestration Software

• Coordinate Complex AI Processes
  Developers can orchestrate multiple steps, agents, loops, and parallel paths in a single workflow, supporting more complex applications than isolated steps.
• Control Data Across Workflow Steps
  Type-safe programs let developers specify how arguments and data pass between steps, providing precise control over workflow behavior.
• Run Workflows Asynchronously
  Asynchronous execution supports fast workflow runs and integration with Python applications, including FastAPI.
• Pause And Resume Stateful Workflows
  Start workflows on demand, pause them, and resume them while maintaining state across the process.
• Include Human Review In Workflows
  Developers can incorporate human-in-the-loop review alongside document processing and other workflow steps.
• Build And Deploy Open-Source Applications
  Developers can use the open-source framework to build applications with control over their implementation, including production and commercial applications.

LlamaAgents

LlamaAgents is a low-code builder for developing and deploying document agents from natural-language instructions. It enables developers and enterprises to build reliable document workflows spanning ingestion, extraction, reasoning, and automation.

• Automate Document Workflows
  Connect document ingestion, extraction, reasoning, and automation in workflows built around documents.
• Build Agents From Natural Language
  Create document agents using natural-language instructions with a low-code builder.
• Deploy Document Agents
  Deploy document agents developed with the builder to support document workflow automation.