# Zipstack/unstract

LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows

Repository: https://github.com/Zipstack/unstract
Canonical: https://ross.abutalabs.com/products/unstract
Homepage: https://unstract.com
Language: Python
License: AGPL-3.0
License Family: copyleft
Topics: ai-agents, data-engineering, document-ai, generative-ai, idp, json-extraction, llm, mcp-server, ocr, pdf-extraction, prompt-engineering, structured-output
Last push: 2026-08-26T16:41:01+00:00

## Health v2 (maintenance only)
Score: 88/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 86, longevity 66
- inputs: {"age_days": 924, "days_push": 7, "days_rel": 13, "gap_med": 0.0, "n_releases_24m": 415}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7172, forks 709 (observed 2026-08-28T04:09:56.392979+00:00)

## What it is
Unstract is an open-source, LLM-driven platform that extracts structured JSON data from unstructured documents such as PDFs, images, and scans using natural-language prompts. It supports deployment as APIs, ETL pipelines, and an MCP server, with features like dual-LLM validation (LLMChallenge) and token-saving extraction modes.

## Use cases
- extract structured data from pdfs with llm
- parse invoices into json automatically
- build an etl pipeline for unstructured documents
- deploy document extraction as an api
- process bank statements without templates
- add document extraction to ai agents via mcp
- reduce llm token costs for document parsing
- automate accounts payable document processing

## When to choose
- you need production-grade extraction from varied document formats without training or templates
- you want API or ETL deployment of document extraction workflows
- hallucination control matters - dual-LLM consensus validation is valuable
- you need connectors to file systems and databases for bulk document processing

## When to avoid
- you only need simple OCR text extraction without structured output
- you want a lightweight library to embed in code rather than a platform to deploy
- AGPL-3.0 licensing is incompatible with your usage
- your documents are already structured (CSV, JSON) and need no LLM parsing

## Facets
- artifact type: application
- maturity: active
- function: ocr, etl, rag, prompt-engineering, mcp, llm-inference, data-science, api-framework, self-hosted
- domain: artificial-intelligence, large-language-models, pdf, developer-tools
- platform: python, self-hosted, cloud
- tags: intelligent-document-processing, idp, document-ai, structured-output, json-extraction, llm-challenge, prompt-studio, etl-pipelines, agentic-ai, pdf-extraction, data-engineering, automation, docker, web-server

## Member repositories
- Zipstack/unstract (main) score 88

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:56.392979+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:39:56.574411+00:00, confidence not recorded.
  - readme: https://github.com/Zipstack/unstract (fetched 2026-08-28T04:09:56.392979+00:00, sha 502c7b6ca6c8)
  - homepage: https://unstract.com (fetched 2026-08-29T08:35:20.746887+00:00, sha 7523b9e8d5cb)
  - site_page: https://docs.unstract.com/ (fetched 2026-08-29T08:35:20.758098+00:00, sha 3b8e138d44cc)
  - site_page: https://unstract.com/pricing (fetched 2026-08-29T08:35:20.756023+00:00, sha 19b4c6be2d27)
  - site_page: https://unstract.com/unstract-editions (fetched 2026-08-29T08:35:20.759777+00:00, sha c4a68faa5aef)
  - site_page: https://unstract.com/ai-accounts-payable-procurement-document-processing/contract-pricing-proposal-data-extraction (fetched 2026-08-29T08:35:20.761948+00:00, sha 27cacae9dbce)
- Data as of 2026-08-30T08:39:29.467469+00:00.