Skip to content
Docuoria

Public preview · version 1.1.3

Your invoices, in a spreadsheet.
No retyping.

Docuoria works inside the AI assistant you already use. Ask for the data in a PDF and get a clean spreadsheet, extracted on your machine, the same way every time.

$ npm install -g @sidub-inc/docuoria.cli$ dotnet tool install -g dotnet-script$ docuoria init

Needs Node.js 20 and the .NET 10 SDK. Also available as a .NET tool. All install steps

Works inside

  • Claude Code
  • GitHub Copilot
  • Cursor
  • Windsurf
  • Codex
  • VS Code AI Toolkit
  • AGENTS.md

What you ask for, what you get

Three things people ask Docuoria to do.

Each one has been run end to end on real vendor invoices and receipts, in two currencies and four date formats, and checked against hand-verified values.

One invoice, one spreadsheet

> Can you get the data out of this invoice? I need it in a spreadsheet.

A CSV with the invoice number, date, and total, and one row per line item.

All your software bills, one sheet

> All my software bills from different companies, in one spreadsheet so I can see what we are spending.

One CSV with a vendor column and a currency column. Amounts in different currencies are never added together.

Exactly the fields your bookkeeping tool imports

> Our bookkeeping tool imports JSON. I need vendor, invoice number, date, currency, pre-tax, tax, total.

One JSON file per invoice with those seven fields, typed as numbers and dates, and nothing else.

All use cases, with the output each one produces

Set up once

Set up once per kind of document. After that, one sentence.

The first time

Your assistant does the work once

It reads every page, works out where each value lives, checks that the amounts add up, and saves the rules. If the document can reasonably be read two ways, it asks you one question.

The next time

One sentence

"New invoice came in. Same as last time please." The engine recognises the document, applies the saved rules, and hands back the file. No questions.

Every month

One running spreadsheet

"Add July’s." The ledger grows without touching earlier rows. Run it twice by mistake and nothing is added twice.

Why the numbers hold

The assistant sets it up. A rule-based engine runs it.

Docuoria splits the job in two: your assistant works out where each value lives, then a rule-based engine reads it the same way every time.

Your AI assistant, once per document type

  • Reads every page of the document
  • Tests each pattern against the real text before using it
  • Runs a rehearsal and checks that subtotal plus tax equals the total
  • Saves the method as a set of rules it can reuse

The Docuoria engine, every time

  • Recognises the document and picks the saved rules
  • Applies the rules with no model involved
  • Never hands back half a result: a run either completes, or tells you which field it could not fill
  • Same PDF, same rules, same output, on every run

How it works, step by step

Privacy

Your documents stay on your machine.

Extraction runs inside the process on your computer. The PDF, the values taken from it, and its file name are never sent anywhere.

Docuoria contacts one service, api.monaiq.com, to validate your licence and to report how many operations you ran. That report carries no document content. You can watch the connections yourself while an extraction runs.

What leaves your machine, and how to check

Stays
PDFs, extracted values, file names, saved rules
Leaves
Licence check and aggregated operation counts

Getting started

Three steps to the first file.

  1. 1

    Install

    Three commands add the Docuoria skill to the AI tools you use.

  2. 2

    Sign in once

    Get a free licence with one command, or paste a key from the marketplace.

  3. 3

    Ask

    Describe the outcome in plain language. The file lands next to your PDFs.

Free for personal and non-commercial use.

Pro, for commercial use, is $9.99 CAD a month with higher limits and a shared template store for teams. The saved rules are the same on both plans.