Open-source PDF editor: PdfCraft tested

PdfCraft is an open-source PDF reader and editor rebuilt from scratch in Rust by the ArtCraft team (the repository is pdfcraft), built around three promises: read any PDF faithfully, append your changes on every save without touching the original bytes, and work offline with no account. As of October 12, 2026 it has 10,384 GitHub stars (2,305 at the seven-app comparison on October 8), with the latest release v0.6.0 shipped on October 11. Its self-ratings come in three flavors: 57% of tracked Acrobat Pro offline features, 94% of must-haves, and 55% for day-to-day replacement of Acrobat — 49% against all of Acrobat Pro, 66% for someone who only needs the essentials. This article separates what genuinely works today from what does not, because a PDF tool’s value depends almost entirely on what you ask it to do.

Key points

Read the three numbers together: 57% is full-feature coverage, 94% is must-have coverage, and 55% is the day-to-day figure. The project’s own phrasing is “every everyday workflow can be completed, but converting and sharing is weak” — not a contradiction, three cuts of the same fact.
The thin areas are four specific things: reliable editing of existing text (especially CJK and right-to-left), OCR beyond Latin script, Office import and export, and signature tokens with online long-term validation. Preflight, PDF/X/UA and full XFA forms are also absent.
Encryption and permissions are ahead of expectations: a complete PDF standard security handler, every revision from 40-bit RC4 to AES-256, user and owner passwords including SASLprep-normalized Unicode ones, crypt filters and attachment-only encryption — and the app honors the document’s own restrictions.

What 57%, 94% and 55% mean

PdfCraft’s status section (October 2026) is unusually explicit: alpha, about 57% of tracked Acrobat Pro offline features and 94% of must-haves; roughly 55% of the way to replacing Acrobat for everyday work (viewing, commenting, filling and signing forms, organizing pages), about 49% against all of Acrobat Pro, and about 66% for someone who only needs the essentials. The method is documented in docs/target-app-parity.md and the ranked gaps live in docs/gaps.md.

Two weeks ago the comparison article recorded roughly half of Acrobat’s offline features and “a third of the work.” The current framing is finer-grained and checkable — the 37 points between 57% and 94% are exactly the features that exist by name but not in full: form authoring depth, editing existing text, some export options.

The honesty is worth noting on its own. The README says page drawing is still done by the hayro crate while the project’s own renderer is being built, and that fuzzing still turns up crashes and hangs on hostile files — each one fixed with a regression test — and that quality has not yet been compared with Acrobat side by side. Projects that print those two sentences in their README are usually worth taking at their word elsewhere.

The price context makes the decision easier than the feature list. Acrobat Pro lists at $19.99 a month in the US (read from Adobe’s plans page via OrcaRouter on 2026-10-07), which is $239.88 a year for a workflow that most people use a few times a week. Against that, the relevant question is not whether PdfCraft is free — it is which of your recurring PDF tasks it fails at. If the answer is “none that matter this month,” the subscription was covering a habit rather than a capability.

What works today

By the project’s own accounting, these areas are good today:

Viewing and search, including correct rendering of complex scripts (covered below), full-text search, and navigation through long documents.
Organizing, combining and splitting: combine any number of PDFs, each file getting a bookmark with its own bookmarks nested underneath; extract selected pages into a new document; split every n pages or before pages you choose. Details survive the trip — links and named destinations are rewired to the copied pages, form fields stay interactive, layers keep their on/off defaults, attachments come along, and every page of a combined document renders pixel-identical to its source. The CLI covers the common cases:

pdfcraft-cli combine report.pdf appendix.pdf --out combined.pdf
pdfcraft-cli extract report.pdf --pages 1,3,5 --out highlights.pdf
pdfcraft-cli split   report.pdf --every 10 --out-dir parts/

Comments: highlights, shapes, ink and stamps, with comments presented as threaded conversations; optional content (layers) toggles and the page re-renders instantly.
Forms: every field with its current value, field highlighting, and checkboxes, radio buttons, lists and signatures drawn as their author designed them, with sandboxed JavaScript.
Passwords, redaction and sanitizing: the full security handler (its own section below) plus redact and sanitize workflows. Sanitizing is the one people underrate — stripping metadata, embedded files and hidden layers before a document leaves your organization is a routine requirement in legal and HR work, and doing it by hand in Acrobat is slow enough that people skip it.
Basic digital signatures — appearance and an integrity record, not token-backed signing.
Printing from the desktop app.
The Accessibility Checker, which checks tagging and structure.
Agent control through the CLI and an MCP server.

Line that list up against your actual day and the answer is sharp. If your day is “receive a PDF, comment, fill forms, merge, sign, send,” PdfCraft does all of it now. If it is “fix a typo in an existing Chinese contract, export print-ready PDF/X, run Preflight,” it does none of it yet.

Two smaller interfaces deserve a mention because they are what people use every hour: navigating long documents (outline, thumbnails, page jumps, search results that land on the right page) and organizing pages like cards on a table. Many free tools bury these behind a “professional” menu. PdfCraft puts them in the same layer as viewing, which matches its choice to list viewing and search first among what is good today — reading is a PDF tool’s primary job.

The “every tool, one keystroke away” framing in the README is also worth noticing. Every action in this application is meant to be reachable from the keyboard, which sounds mundane until you have filled in forty form fields by hand. Form work is where a PDF tool earns its keep in administrative and back-office roles, and it is also where sandboxes matter: PdfCraft runs form JavaScript in a sandbox, so opening an untrusted document does not hand a stranger script execution inside your machine. Combined with the permission honoring described below, that is the security posture a legal or HR team actually needs from a free tool — and it is the part most free PDF utilities skip.

Thin and missing

The project’s own list of shortfalls, item by item:

Reliable editing of existing text is the biggest gap, especially CJK and right-to-left scripts. For Chinese-language users this is the painful one: Acrobat’s most common workflow — fixing a typo, changing an amount — cannot be relied on.
OCR is limited to Latin script. There is no Chinese OCR for scanned documents today.
Office import and export are missing, and Word export loses layout — the project lists Office-to-PDF plus a layout-preserving Word export as the very next item.
No signature tokens or online long-term validation, so compliance-grade signing is out of reach.
No Preflight, no PDF/X, no PDF/UA, which closes the door on print and archival workflows.
No full XFA forms — traditional AcroForms only.
The web build cannot print, and physical printer fidelity is unverified.
No side-by-side comparison with Acrobat has been done.

One architectural fact explains part of this and is easy to miss: page rendering still borrows the third-party hayro crate while the project’s own renderer is built. That is why the typography section below looks strong while editing existing content remains the top gap — replacing the renderer is scheduled later.

Encryption and permissions

This is the part of PdfCraft worth its own section. It implements the PDF standard security handler completely: every revision from 40-bit RC4 to AES-256; user and owner passwords, including Unicode passwords normalized with SASLprep; crypt filters; and attachment-only encryption.

More important is that it respects the document’s own restrictions. A document restricted by its author shows a clear notice, PdfCraft honors the permissions, entering the owner password lifts them, and edits to encrypted documents are saved encrypted under the same keys. That is more than “it opens encrypted PDFs” — plenty of tools ignore permissions outright.

The Document Properties dialog shows the whole security picture: the encryption method, which password opened the file, and each permission. It also lets you edit title, author, subject and keywords, and lists the fonts the document uses and whether each is embedded, plus the PDF version, page size, tags, fields, layers and attachments.

Typography and rendering

Reading fidelity is where PdfCraft separates itself from tools that merely open PDFs: kerning and ligatures, right-to-left and complex scripts, vertical CJK, color emoji, shadings, blend modes, soft masks and optional content. The README’s example page sets Arabic, Hebrew, Devanagari, Thai, Greek, Cyrillic, Chinese, Korean, IPA, Armenian, Georgian and Tamil samples side by side, with vertical Japanese in the right margin.

None of that is decoration. With scanned documents, government paperwork and Asian-language PDFs in the wild, “can display” and “displays correctly” are two different products — a wrong soft mask or blend mode makes a page look broken. PdfCraft files all of it under “Faithful,” the first of its three headline qualities, which tells you where the team thinks the product’s value is.

The Document Properties panel doubles as the other half of that story. It tells you which fonts a document uses and whether each is embedded, which is the question behind “why does this PDF look different on my machine” — a missing font is the single most common cause of layout drift, and a tool that surfaces the answer saves an afternoon of guessing. It also lists the PDF version, page size, tags, fields, layers and attachments, so when a document misbehaves you can see whether the problem is in the structure rather than in the content.

Installing and first run

Downloads live at getartcraft.com/apps/pdfcraft; the GitHub release page lists every format with a SHA256SUMS.txt checksum list. Windows has MSI installers and portable zips for x64, arm64 and x86, all code-signed. macOS gets a universal DMG, signed and notarized, plus a signed CLI zip. Linux gets an AppImage with .zsync self-update, a Flatpak, .deb, .rpm and a tarball. FreeBSD has a tarball, every release ships a WebAssembly build you can self-host on any static server, and Japanese fonts come from the optional craft-fonts input that release builds always embed.

The CLI is usable on its own, so batch work needs no desktop install at all — and for a first session, running the three combine/extract/split commands against a copy of a real document shows more about the tool’s care than any screenshot: whether form fields stay interactive, whether bookmarks survive, whether the merged page renders identically. Pixel-identical rendering of combined pages is a claim worth testing yourself, because it is the kind of promise that is easy to make and easy to break. One more habit is worth building while the tool is new: keep a copy of the original document before every merge or redaction pass, because saving appends rather than overwrites — that habit is nearly free here, and it turns an accidental overwrite into a question you can answer by diffing two logs.

For teams, the security documentation covering the threat model, parser hardening and vulnerability reporting is material you can hand to whoever signs off on document handling — a rare asset for a free tool that processes contracts and ID documents.

Command line and agents

The CLI covers the three common bulk operations shown above, and the engine, CLI and app are all open source. The agent surface matches its siblings: a command-line tool, an MCP server and a control channel dispatch through the same entry point, so an agent can comment, fill forms, redact, sign and merge or split in batches. For teams handling dozens of contracts a day, the pipeline “read the PDF → extract fields → fill forms → merge → export” is scriptable today, entirely offline, with no account and no telemetry.

The append-only save model matters most in automation: when a script goes wrong you can return to the previous state, and you can diff exactly what changed across a run.

For agent work, the useful framing is that the three properties of the product — faithful reading, append-only saving, offline operation — are exactly the three properties automation needs. An agent that misreads a page produces a wrong edit you can inspect; an agent that overwrites a file destroys the audit trail you need. Here the failure modes are recoverable by design. The roadmap also names the reported bugs and the Chinese interface font as near-term items, which is a reasonable signal of where the team thinks the rough edges are for this audience.

What the roadmap queues next

The project’s own ordering is more informative than a feature matrix, because it is written in the sequence it intends to attack the gaps: Office-to-PDF and a Word export that keeps layout first, then the reported bugs and the Chinese interface font, then a fidelity harness against Acrobat and against their own renderer, then editing existing content, and only after that the professional workflows — OCR, Office, Preflight, signatures and accessibility tagging — plus 1.0 polish.

Read in order, that sequence says three things a buyer should know. Chinese text editing is not next; it comes after the Office conversion and the fidelity harness. The Acrobat comparison is being built as a test harness, which means the 49–55% figures are expected to move on their own once real comparisons exist. And 1.0 is several milestones away, so “alpha” here means alpha in the ordinary sense, not as a polite label for a finished product.

PdfCraft vs Stirling-PDF

The free PDF field in 2026, with facts checked on 2026-10-12:

Dimension	PdfCraft v0.6.0	Stirling-PDF v3.1.0	Acrobat Pro
Price	free, MIT / Apache-2.0	free, AGPL-3.0	$19.99/month (US, 2026-10-07)
Shape	native desktop app + CLI + WebAssembly	self-hosted web app (Docker)	desktop app + cloud services
GitHub stars	10,384	94,046	—
Editing existing text	weak (CJK and RTL especially)	weak (reviews call it behind Acrobat)	strong
Encryption and permissions	complete across all revisions, honors restrictions	limited	strong
Forms	AcroForm plus sandboxed JS; no XFA	supports XFA	complete
OCR	Latin script only	multi-language route built in	complete
Office conversion	none	limited	yes
Collaboration and sharing	local-first	fine for self-hosted teams	Adobe cloud

Stirling-PDF has nine times the stars of PdfCraft and leads on OCR and self-hosted collaboration; the shared verdict in third-party reviews is that its editing lags Acrobat. PdfCraft bet the opposite way — encryption, forms, typography fidelity and merge/split that loses nothing, which it does about as cleanly as Acrobat does. Neither tool is an either/or choice: Stirling-PDF fits bulk OCR and a shared team service, PdfCraft fits a desktop workflow that needs document structure changes, permission handling and offline work. For PDF/A archival or compliance signing, neither satisfies the requirement today, and that tier still needs commercial software.

The practical split comes down to where the document lives and who needs it. If your team wants a URL that anyone on the network can use to convert, merge and OCR files without installing anything, Stirling-PDF is the product built for that, and its Docker deployment is a weekend project. If you work on documents that arrive with restrictions, need a desktop tool that opens instantly without a server, and care about whether a merge preserved the form fields and the bookmarks, PdfCraft is the one built for that. Running both costs nothing, and the overlap — viewing, merging, basic forms — is exactly where they agree.

Who should try it, who should wait

People who live in PDFs all day — contracts, reports, scanned forms: worth a trial. Viewing, search, comments, forms, merge and split, encryption, redaction and signing are all present, and it works the moment you install it. Acrobat Pro costs $19.99 a month; saving that costs nothing.
Workflows that edit existing text, Chinese especially: wait. Editing text is the top-ranked gap, and the roadmap puts it behind Office conversion and the fidelity harness.
Print, archival and compliance signing: stay on commercial software. Preflight, PDF/A and signature tokens are hard requirements, not preferences.
Developers and automation: usable now. The CLI plus MCP suits field extraction, redaction and merging, and the append-only save model makes pipelines auditable and reversible.

One practical note for anyone evaluating it: install it beside your existing Acrobat rather than instead of it. The test that answers most of the doubt is to take three real documents from your last week — one to annotate and send, one form-heavy document, one that arrives encrypted with owner restrictions — and run them through PdfCraft end to end. If all three work, the remaining gaps are the ones you personally do not touch, which for most people is a good trade for $19.99 a month.

PdfCraft belongs to a bigger picture: our seven-app comparison weighs all of ArtCraft’s Adobe alternatives. The GitHub repository updates daily; downloads live at getartcraft.com/apps/pdfcraft, with code-signed Windows installers, a signed and notarized macOS DMG, AppImage, Flatpak, deb, rpm and tarballs for Linux, a FreeBSD tarball and a self-hostable WebAssembly build. The CLI is usable on its own for batch work, and the project’s security documentation covers the threat model, parser hardening and how to report a vulnerability — material worth handing to whoever signs off on your document handling.

FAQ

Is PdfCraft free for commercial use?

Yes. MIT or Apache-2.0 dual license, no account, no cloud, no telemetry, fully offline; the engine, CLI and app are all open source. There is no licensing cost for commercial use.

Can it edit text inside a PDF?

Partly, and this is today’s biggest gap. The project describes reliable editing of existing text as unavailable, especially CJK and right-to-left scripts, and Office import and export is still missing with layout loss on Word export. To change text today, edit the source document and export to PDF, or keep using commercial software.

Does it handle encrypted PDFs?

Yes, and more correctly than most free tools. It implements the PDF standard security handler completely — every revision from 40-bit RC4 to AES-256, user and owner passwords with SASLprep normalization for Unicode passwords, crypt filters and attachment-only encryption — honors the document’s own permission restrictions, and saves edits to encrypted documents encrypted under the same keys.