# Unstract

An LLM-driven platform that extracts structured JSON from unstructured documents such as PDFs, scans and images, deployable as an API or an ETL pipeline.

Unstract uses LLMs to extract structured JSON from documents -- PDFs, images, scans and more -- letting a user define what to extract with natural-language prompts in its Prompt Studio, then deploy the result as an API or an ETL pipeline. It is built for teams in finance, insurance, healthcare and KYC/compliance, among others, and plugs into any LLM provider, including OpenAI, Anthropic, Bedrock or Ollama.

Its API Deployment mode returns JSON over REST for a submitted document, its ETL Pipeline mode pulls documents from a folder, processes them and loads the result into a warehouse, an MCP Server connects Unstract to AI agents such as Claude, and an n8n node drops extraction into existing automation workflows.

It runs locally via Docker Compose with a single setup script, supports document formats spanning PDF, Word, spreadsheets, presentations and common image types, and encrypts adapter credentials with a backup encryption key that must be saved securely, since losing it makes existing adapters inaccessible.

Website: https://unstract.com