Product
DocShift
Private PDF tools that run on your own device, as a web app, a desktop app and a CLI.
- Started
- 25 June 2026
- Status
- Live
- Signal
- PDF tools
The question
Can a PDF tool feel boring in the best possible way?
Most PDF tools work, but each one adds a small frustration: upload limits, accounts, watermarks, slow processing, too many buttons, or the hesitation that comes with uploading a private document somewhere you do not fully trust. I wanted the version I could recommend without thinking twice.
Context
DocShift came out of PeerDrift.
PeerDrift answered one question: can you move a file between two devices without handing it to a server you do not trust? Once that worked, people were not only moving documents around, they were trying to fix them. Merge these. Compress this so it fits the upload limit. Rotate the scan. Turn it into Word.
The same constraint carried over. If the file never has to leave the device to be shared, it should not have to leave the device to be edited either. The privacy model, the no-account rule, and the Go and Wails desktop shell all came straight from PeerDrift, which is why DocShift got to a working desktop build so quickly.
Building it
Everything runs client-side. No account, no upload, no watermark, no external API for the document tools themselves.
What it does today:
- Compress with Safe, Balanced and Smaller presets
- Merge, split, reorder, rotate, delete and extract pages
- Add page numbers and watermarks
- OCR scanned pages so they become searchable, with plain text export
- Prepare fillable forms by detecting labels, blank lines and checkbox cues
- Password-protect a PDF, or unlock one when the password is known
- Convert between PDF, JPG, Word, PowerPoint and Excel where a browser-side conversion actually produces something useful
It runs in three places: the web app, a Go and Wails desktop shell around the same frontend, and a CLI published as @tonycletus/docshift for scriptable work.
The engineering question underneath is how much of a document workflow can stay local before a server becomes necessary. Some conversions are honest about their limits, which is why the tool only offers the ones that produce a result worth keeping.
What surprised me
The people using it care more about correctness than speed. A CV, an invoice, a school document or a legal paper looks ordinary from the outside, but the user usually has one shot at sending it, so a silently mangled conversion costs them more than a slow one.
What I learned
Preview and failure states did more for trust than any feature in the list. Showing the output before the download, and saying plainly when a conversion will not be clean, is what makes a local-first document tool usable by someone who cannot check the result themselves.