Paperless-AI is an automated document analyzer for Paperless-ngx, the popular self-hosted document management system. It sends newly archived documents to a language model and automatically generates titles, tags and other metadata, so a growing paperless archive organizes itself. The project has about 6,000 stars on GitHub, is MIT licensed, and its Docker image has been pulled heavily by the self-hosting community.
Paperless-ngx already OCRs and stores your scans; Paperless-AI adds the understanding step. It connects to the provider of your choice — OpenAI’s API, a local Ollama instance, DeepSeek, Azure, or any OpenAI-compatible endpoint — analyzes each document’s text, and writes back an AI-generated title and tags. Running everything through a local Ollama model keeps private documents off third-party servers entirely. The author notes in the repository that the current codebase is in maintenance while a ground-up rewrite is underway, but the existing Docker deployment remains the version the community runs.
Setup is deliberately boring: one additional container next to Paperless-ngx, an API key or a local Ollama endpoint, and options controlling which tags the AI may apply and how titles should be formatted. Users can review what the AI changed, which keeps the automation trustworthy for archives that took years to scan. For households and small offices drowning in scanned mail, it removes the least enjoyable step of going paperless — deciding, document by document, what everything should be called.
Free. Paperless-AI is open source under the MIT license; you pay only for any hosted LLM API you choose to connect, and nothing when using local models.
Paperless-ngx users whose archives have outgrown manual filing and who want an AI to keep titles and tags consistent automatically.