Unstract is an open-source platform for extracting structured data from unstructured documents with LLMs. Give it invoices, contracts, forms or reports as PDFs and images, describe the fields you want, and it returns clean JSON โ ready to deploy as an API or plug into ETL pipelines. The project has about 7,300 stars on GitHub and is actively developed in 2026.
Unstract replaces brittle template-based OCR rules with LLM-driven extraction: a no-code Prompt Studio lets you define extraction schemas and test them against sample documents, and its document preprocessing companion, LLMWhisperer, cleans up scans, tables and handwriting so the models see the best possible input. Workflows chain extraction, human review and validation steps, and finished pipelines deploy as APIs or run inside your data stack. It supports models from OpenAI, Anthropic, Google and open-source providers, includes challenge/validation runs to measure extraction accuracy before going live, and exposes an MCP server so AI agents can call extraction as a tool. Typical uses include invoice and receipt processing, KYC document checks, insurance claims and financial statement extraction. The codebase is AGPL-3.0 licensed and can be self-hosted with Docker, while the same team offers a managed cloud for teams that do not want to run infrastructure.
Unstract’s open-source edition is free to self-host under the AGPL-3.0 license. The managed Unstract Cloud and enterprise plans are paid, with usage-based pricing for hosted extraction and support.
Unstract is for data and operations teams that process large volumes of documents โ invoices, contracts, claims, forms โ and want reliable structured output without maintaining fragile OCR templates.