Paperless GPT is an open-source companion service for Paperless-ngx that uses LLMs and vision models to organize documents automatically, published on GitHub with about 2,800 stars. It generates AI-powered titles, tags and correspondents for scanned paperwork and performs LLM-based OCR where needed.
Paperless-ngx archives documents, but naming and tagging them is manual work. Paperless GPT watches the archive, sends document text and images to a language model of your choice, including local models via Ollama, ChatGPT or Mistral, and writes back structured metadata: descriptive titles, consistent tags and suggested correspondents. A manual mode and a review UI let you approve suggestions before they are applied, and an experimental OCR path uses vision LLMs to extract text from difficult scans. The project is MIT licensed, distributed as a Docker image, and actively maintained in 2026. It is a separate project from Paperless-AI, with a different approach and maintainer.
Paperless GPT is free and open source under the MIT license for self-hosting. For users in Germany, Austria and Switzerland who prefer not to run it themselves, an official managed-hosting option through a partner provider is offered, so both free and paid paths exist.
Paperless GPT is for Paperless-ngx users, home archivers and small offices who want their document archive organized by AI instead of by hand.