Messy documents in.
Clean, reviewed data out.
Point it at a folder of PDFs, scans, Word and Excel files. It works out which ones matter, pulls out the fields you define, sends anything uncertain to a human, and exports one standardized spreadsheet, for any kind of document.
How it works
Six stages, one screen each. Every stage is visible and correctable.
- 01
Ingest
PDFs, scans, Word, Excel, images. Per-page OCR fallback for scanned or broken-font pages.
- 02
Triage
Your plain-English rules decide what's in scope, so irrelevant attachments never reach extraction.
- 03
Extract
A local LLM fills your fields, with a confidence score and a reason for every blank.
- 04
Review
Only uncertain fields need a human. Side-by-side original, keyboard-driven approve & next.
- 05
Export
One standardized Excel/CSV per project: approved rows, your columns, your labels.
- 06
Fulfil
Turn an approved request into a task: invite vendors, draft emails, track delivery and invoice.
Start from a template
Ready-made fields and relevance rules for common document types. Edit anything before you create.
Private by default
Runs on your own machine with Ollama. Documents never leave your network.
One app, many verticals
Each project has its own rules, fields and sources. Add a new document type without writing code.
Portable templates
Export a project's template as JSON and import it anywhere, or start from a ready-made one.
Built for messy inputs
Image-only pages, broken PDF fonts, Word checkbox forms: the edge cases are handled, not ignored.