Validate input
Confirm the file type and keep untrusted documents within your normal security controls.
Document workflow
Turn incoming documents into usable text.
Combine attachments, extract embedded PDF text and use OCR when a scan needs it. The output can feed your own search, classification, review or summarization system.
A practical intake sequence
Keep a human review step for consequential documents. Extraction makes text accessible to a workflow; it does not guarantee the source or interpretation is correct.
Prepare the packet
Merge related PDFs, split unnecessary pages or rotate pages before extraction.
Read the text
Use PDF-to-text for embedded text. Use OCR for scanned images and image-only documents.
Send it onward
Pass the result to your own archive, search index, agent or review queue.
Tools for the job
OCR availability: PDF text extraction is supported. Image OCR depends on the server OCR engine and reports clearly when that capability is unavailable; test it with a representative non-sensitive document before designing a production dependency around it.
Common questions
Is PDF-to-text the same as OCR?
No. PDF-to-text reads text already embedded in a PDF. OCR attempts to recognize characters in a scan or image.
Can I send extracted text directly to an LLM?
You can, but first consider document sensitivity, prompt size and your model provider’s data practices. The token counter helps estimate input size.
Can an agent use these tools?
Yes. Authenticated agent routes are documented for supported PDF operations, and the workspace exposes the wider document toolkit.
Start with one representative document.
Verify extraction quality, then connect the working path to your agent or application.