Getting Started

What is Extend?

Extend is your platform for building, iterating on, evaluating, and deploying AI-powered document processing infrastructure. We support a variety of tasks like document extraction, splitting, classification, and parsing. These tasks are performed by Processors, which can then be orchestrated within Workflows to handle your document processing use cases.

What do you want to do?

Extract structured data from documents Recommended
→ Quick Start: Build your first extraction processor

Split multi-page documents into organized sections
→ Configure a splitter

Just parse documents into clean, structured text
→ Parse documents

Categorize documents into different types
→ Configure a classifier

Orchestrate multiple processors into complex document pipelines
→ Create a workflow

Test and measure how well your processors perform
→ Overview of evaluation sets

Coding with an agent?

Use these raw markdown files to seed coding assistants with Extend platform context:

  1. Download one of these files and save it in your project root as CLAUDE.md or AGENTS.md.
  2. Run your coding agent from that same project root so it can discover the file.
  3. If your agent does not auto-load root guidance, explicitly reference the file in your prompt (for example: “Use ./CLAUDE.md for project instructions.”).
  4. For Codex/agent-style tools, keep AGENTS.md in the working directory root where commands are executed.
  5. For Claude Code, keep CLAUDE.md at repository root and start the session from that directory.
  6. In web LLM chats (for example ChatGPT, Claude, Gemini), upload CLAUDE.md or AGENTS.md as an attachment at the start of the conversation.
  7. If attachments are not available, paste the file content into your first prompt and ask the model to follow it for the rest of the session.