docling-project/docling
Extract tables and text from complex or scanned PDFs with Docling
An open source document parser that understands page layout, reading order and tables, with OCR for scanned files. It outputs Markdown or JSON and can be connected to agents through its MCP server.
When to use it
Use it when simple converters lose the structure of invoices, financial statements or research reports and you need tables kept intact.
What it can access
- Runs code
- Reads or writes files
Before you install it: Runs on your own machine and downloads AI models on first use; documents are processed locally unless you choose a remote model. It can be heavy on CPU and memory for large batches.
Open-source tools often need access tokens for your apps. Wize reads the code, keeps the access as narrow as the job allows, and sets it up so a person approves anything that goes out.