I.2 The Documentize idea: one app = one job
The promise in four verbs
Imagine a drawer of good kitchen knives instead of one Swiss army gadget. You grab the one you need, use it, put it back. That is the whole idea, and it comes down to four verbs:
Open. Drop. Run. Download.
You open the page for the job you have, drop in your file, press the button, and download the result. The apps in the classic family say it themselves in their "How it works" panel: select files, press the button, wait for completion. Wait times depend on file count and size; the panel says it takes from ten seconds to several minutes.
There is nothing to install. For most one-off jobs you do not have to sign in either, although how many runs you get per day depends on whether you are signed in and on your plan (see Limits, plans and the daily allowance).
One job, named on the page
Each app does a single, nameable thing:
| App | Its one job |
|---|---|
| Merger (page, chapter) | Many files in, one PDF out |
| Compressor (page, chapter) | Make the file smaller by softening pictures |
| Unlock (page, chapter) | Remove a password you already know |
| Searchable (page, chapter) | Add an invisible text layer to a scan |
| Ask PDF (page, chapter) | Answer questions, with a page citation for every claim |
Notice what is missing: Compressor does not also straighten pages, and Unlock does not try to guess passwords. The book is blunt about it where the product is blunt about it. Unlock "is not a cracker"; Optimize "does not reduce visual quality"; Protector sets one password to open, not a menu of permissions.
This narrowness is a feature. When an app does one thing you can predict it. The page of each chapter answers the same twelve questions, from "In one sentence" to "Try it", because the apps are simple enough to deserve that.
Results feed the next app
The output of every app is an ordinary file. Your original is never modified; every run produces new files. So any result can be dragged into the next app.
flowchart LR A[Scans + Word + photos] --> B[Merger] B --> C[Deskew] C --> D[Searchable] D --> E[Compressor] E --> F[Protector] F --> G[Send]
That chain turns the opening mess of the problem into a single tidy pack: Merger, Deskew, Searchable, Compressor and Protector, each one click away. Each arrow is a download and an upload, or, if you automate, one command after another (see Beyond the browser).
What "one app, one job" does not mean
- It does not mean the apps are identical. Some are exact and repeatable, such as Splitter and Add TOC. Some are AI-driven and give a different answer each run, such as Table of Contents AI, the AI Illustrator and SVG Maker. The chapters say which is which; the quality model explains why it matters.
- It does not mean every app is gentle. Compressor deliberately lowers picture detail. Deskew rebuilds each page as an image. Order matters; Part III gives the golden rules.
- It does not mean every app exists. Two apps appear in the plan, PDF Redactor and Compare PDFs, but are not available as working apps. Their chapters say so and offer a manual route. Three developer apps are retired (Release Notes, Fork Sync, Personal Engineer). This book would rather tell you than hide it.
Why the idea is worth having
- Learn once, use everywhere. Every app shares the same flow, the same kinds of options and the same result page. See Part II.
- Small blast radius. A wrong click in a single-purpose app changes one thing and leaves your original intact.
- Chains. Because the pieces are small and honest about their state, you can combine them into recipes (Part XII).
- Same engine, other doors. The same apps are available as command-line plugins, an AI-assistant (MCP) server and a web API, so a chain you do by hand today can run unattended tomorrow.
Next
You have met the idea. But there are two addresses, and the first question is always "which one do I open?" The answer fits on one page: Two sites, one product family →