ParseCrate
Your documents, ready for your applications
Open-source, self-hosted document processing API.
Turn PDFs, Office files, and other documents into structured content on your own server. Powered by Docling and docling-serve, with OCR, table recognition, and an API for your applications.
Open source · MIT licensed
Try in Google Colab No local installation. Runs in Google's environment.
The Self-Hosted AI Builder's Guide Practical guidance for deployment, security, backups, and everyday operation.
Documents → structured content
report.pdf · a document you choose
Markdown, JSON, or HTML for your application
A focused service for your applications
Convert documents for search, retrieval, archives, and downstream applications. Keep the parsing service on infrastructure you operate, and check extraction quality before relying on it.
Submit supported files or source URLs to conversion endpoints. Choose Markdown, JSON, or HTML output through Docling’s conversion options.
Extract text and structure from documents, including supported scanned inputs. Check reading order, OCR accuracy, and tables against the original.
Use synchronous conversion for suitable jobs or asynchronous conversion for longer tasks. Your application manages requests, results, and errors.
Hierarchical and hybrid chunking endpoints prepare document passages for retrieval-augmented generation (RAG). Add your own embeddings, storage, and retrieval logic.
CPU images support amd64 and arm64. The optional :cuda image supports NVIDIA acceleration on amd64 with compatible host drivers and the NVIDIA Container Toolkit.
Enable the playground when you want an interactive conversion interface. It is disabled by default; persistent storage keeps configuration and model files across restarts.
Images are automatically built and published through public GitHub Actions workflows. Powered by Docling, docling-serve.
Start with one document
Prefer to try it first? Open the Google Colab demo. Documents submitted there are processed in Google's environment.
This CPU example binds the API to the Docker host. Models are included in the image and load on startup.
For access from other computers, follow the maintained HTTPS, authentication, and network-exposure instructions.
docker run \
--name docling \
--restart=always \
-v docling-data:/var/lib/docling \
-p 127.0.0.1:5001:5001 \
-d hwdsl2/docling-serverCheck startup with docker logs docling. Wait for “ParseCrate document parsing server is ready”.
parse_api_key="$(docker exec docling docling_manage --getkey)"Fresh persistent installations automatically generate and save an API key. ParseCrate uses the X-Api-Key header.
curl http://127.0.0.1:5001/v1/convert/file \
-H "X-Api-Key: $parse_api_key" \
-F "files=@document.pdf"Replace document.pdf with your file. Run the request in the same shell as the key command.
Part of a larger toolkit
ParseCrate is also available as part of Self-Hosted AI Stack. Deploy local models, chat, document processing, voice, and AI tools together, or choose a smaller configuration for your needs.
Explore Self-Hosted AI StackConnect the services you need
ParseCrate extracts document content. EmbedCrate can turn that text into embeddings and optionally rerank retrieved passages. Your application connects the services.
Explore EmbedCrateThe ParseCrate project is open source under the MIT License. You provide hardware or cloud hosting and pay associated costs. Models, dependencies, and external providers retain their own licenses and terms.
No. Deploy ParseCrate on your own infrastructure and send documents to that deployment. This website does not offer a hosted document-conversion service.
It handles document conversion and chunking. You still need an embedding model, vector storage, retrieval logic, and an application to connect them.
Use the included or pre-cached models with DOCLING_LOCAL_ONLY=true. Submit local files for offline conversion; URL inputs still require access to their source. Disable aggregate usage counts separately with DOCLING_DISABLE_USAGE_COUNTS=1.
A separate :cuda image supports NVIDIA GPUs on linux/amd64. CPU operation is available on amd64 and arm64. Check the maintained setup guide for current driver and toolkit requirements.
The image includes optional aggregate usage counts.
See the usage-count documentation;
set DOCLING_DISABLE_USAGE_COUNTS=1 to disable them.
Bring your documents. Keep control of where they are processed.
View on GitHub