Wordgard review 2026. Honest scoring on architecture, ProseMirror pedigree, collaboration, accessibility, maturity, license — and who Marijn Haverbeke's new editor library is for.
Wordgard is one of the most interesting developer releases of 2026: a ground-up, open-source rich-text editor library from Marijn Haverbeke, the author of ProseMirror and CodeMirror, that rethinks ProseMirror's ideas with a cleaner change model and CodeMirror 6's facet architecture. Judged as what it is — infrastructure engineers build editors on — it is promising and unusually well-conceived, but genuinely early: v0.1, unproven, with rough docs and known mobile issues. It is MIT-licensed and free, and it produces no content or posts. Watch it if you build software; don't expect it to make anything you publish.
Most write-ups of Wordgard will describe it as "a new rich-text editor for writing and editing," which is close enough to be misleading. Wordgard is not an app you write in. It is an open-source JavaScript library that developers use to build a rich-text editor inside their own product — the same category as ProseMirror, CodeMirror, Lexical, or Slate. That framing matters, because it decides who should care and how to score it.
What makes it worth a serious look is its author. Marijn Haverbeke built ProseMirror and CodeMirror, two of the most respected editor projects on the web, so a brand-new library from him is not a hobby experiment — it is his considered take on how he would do it again, nearly a decade on. Released as version 0.1 on July 2, 2026, Wordgard keeps ProseMirror's premise of a semantic, schema-based document while breaking compatibility to change the internals: a simpler delta-style change model, node and mark types reusable across schemas, CodeMirror 6's facet-based extension system, and its own selection handling built on beforeinput events.
This review scores Wordgard as a developer library, which is the only fair way to judge it. I will not mark it down for not generating video or scheduling posts, because it never claimed to — its own site advertises "0% AI." But I will be clear about the boundary, because "in-browser editor for writing and editing" reads, to a non-developer, like a finished writing tool, and it is not one. The honest read: a thoughtfully designed foundation with excellent pedigree, held back for now only by being brand new.
Wordgard is an open-source library for building in-browser rich-text editors. A developer defines a schema — the exact content types the editor supports, such as text, tables, nested lists, and figures — and Wordgard supplies the document model, a functional-style programming interface, accessibility (screen readers, keyboard, mobile), right-to-left support, collaborative multi-user editing, and pointer/keyboard selection handling. It is deliberately not a drop-in WYSIWYG widget; the point is precise control over what an editor allows, on a foundation you can extend. Architecturally it is a redesign of ProseMirror's ideas incorporating lessons from the CodeMirror 6 rewrite: a delta-based change representation (keep / replace / update) in place of ProseMirror's "steps," schema types that compose across editors, a facet-based system for plugin dependencies, relaxed content constraints with programmatic corrections, and beforeinput-driven input rather than DOM diffing. It ships under the MIT license on what Haverbeke calls an "abundance" funding model. Version 0.1 is explicitly an early, unproven release — documentation is rough, mobile-browser touch-selection issues (it can't fully take over selection without breaking the native context menu) have already surfaced, and he expects to remain on 0.x for at least a year. Treat specifics here as a snapshot of a fast-moving project.
The clear fit is a developer building a custom editing surface — inside a CMS, a docs or notes tool, a knowledge product, or any app where users write structured content — who wants precise schema control, strong accessibility, and collaborative editing on an open, self-hostable base. Teams already comfortable with ProseMirror who want to track where its author is taking the ideas next are the natural audience, as are engineers who value the MIT license and the freedom to own their editor stack. It is a poor fit for anyone who is not a developer, since it is a library rather than an app, and it is the wrong tool entirely for someone whose real goal is producing and publishing content — there is nothing in it for that job.
| Dimension | Score | Why |
|---|---|---|
| Architecture & design | 4.5 / 5 | A clean rethink: delta-style change model, composable schema types, facet-based extensions, custom selection handling. Genuinely thoughtful. |
| Author pedigree & credibility | 5.0 / 5 | By the creator of ProseMirror and CodeMirror — about as strong a track record as an editor library can launch with. |
| Content control (schema-based documents) | 4.5 / 5 | Precise control over supported content types, with structured docs, tables, nested lists, and figures. |
| Collaborative editing | 4.0 / 5 | Real-time multi-user editing is built in, and the delta model is designed to support operational transformation cleanly. |
| Accessibility & internationalization | 4.0 / 5 | Screen-reader, keyboard, and mobile support plus right-to-left handling are first-class, not afterthoughts. |
| License & openness (MIT) | 4.5 / 5 | Free, MIT-licensed, self-hostable, on an "abundance" model that prioritizes wide adoption with no lock-in. |
| Maturity & stability | 2.5 / 5 | Version 0.1, self-described as unproven, with rough docs and known mobile-browser issues; expected to stay pre-stable for a year+. |
| Documentation | 2.5 / 5 | Acknowledged as rough at release. Enough to explore the design, not yet enough to ship a production editor comfortably. |
| Content production & publishing | 1.0 / 5 | Out of scope by design. No generation, no images or video, no scheduler, no publishing — the site advertises "0% AI." |
Wordgard is free and open source under the MIT license, so there is no price tag on the library itself — and for a project of this design quality, that is a genuinely good deal for developers. Haverbeke frames the funding as an "abundance" model, prioritizing broad, unrestricted adoption over licensing complexity, which is consistent with how ProseMirror and CodeMirror have been stewarded.
The honest cost is not money; it is engineering. Adopting Wordgard means a developer builds and maintains an editor on top of it, and at version 0.1 that also means absorbing the risk of a moving target — rough docs, API churn across 0.x releases, and edge-case bugs like the mobile touch-selection issues already reported. For a team that wants an editor it fully controls and is willing to track an early project, that trade is fair. For anyone who needs a settled, drop-in library today, the "free" price does not offset the maturity gap, and a battle-tested option like ProseMirror or Lexical is the safer call for now.
The framing to keep straight: this is infrastructure pricing, not product pricing. There is no hosted app, no seats, and no content output to pay for — because Wordgard is a building block, not a finished tool.
| Use case | Fit | Why |
|---|---|---|
| Building a custom rich-text editor in a product | Strong | This is Wordgard's entire purpose, and its schema control and architecture are built for exactly it. |
| Structured-document editing (tables, lists, figures) | Strong | Schema-first design makes precise structured content a first-class concern. |
| Collaborative, self-hosted editing you own | Strong | Built-in real-time collaboration plus the MIT license and self-hosting give you an editor stack with no lock-in. |
| Replacing or evaluating ProseMirror for a new build | OK | Compelling as its author's next-generation take, but its v0.1 status argues for caution on anything production-critical. |
| A non-developer who just wants to write | Weak | It is a library, not an app; there is no editor to open without building one first. |
| Producing social media content | Weak | No generation or publishing at all — it is the editing layer, not a content maker. |
| Multi-platform content distribution | Weak | There is no scheduler or publisher; distribution is entirely out of scope. |
Wordgard and Kompozy are not rivals; they are different layers, and reviewing one against the other would be a category error. Wordgard is where writing gets *edited* — the foundation a developer builds an editing surface on. Kompozy is where writing becomes *content* and gets *published*. Scoring Wordgard on content production would be unfair to a library that never set out to make posts, which is why that dimension is noted as out of scope above.
The useful way to see the two is in sequence. A draft — typed into a Wordgard-powered editor, or any editor — is an input. Hand that input to Kompozy and it fans one source into 18 output formats: persona and avatar video, clipped shorts, carousels, quote graphics, infographics, blogs, newsletters, and platform-native text posts, all held to one brand voice through a Persona Brief, then scheduled and published across nine platforms plus email and blog from a single queue. Kompozy runs that generation on managed Claude and OpenAI models, so there is nothing to build or operate. If your job is shipping software with a great editor in it, Wordgard (or ProseMirror) is the tool. If your job is turning what gets written into a week of on-brand posts everywhere, that is the specific gap Kompozy fills — and no editor library, however well made, closes it.
Wordgard is an open-source JavaScript library for building in-browser rich-text editors, released as version 0.1 on July 2, 2026 by Marijn Haverbeke, creator of ProseMirror and CodeMirror. It is a foundation developers build editing surfaces on — not an app you write in — and reworks ProseMirror's ideas with a cleaner change model and CodeMirror 6-style extensions.
For a developer building a custom editor who wants a thoughtfully designed, MIT-licensed foundation and is comfortable tracking an early project, it is a very promising choice with rare pedigree. If you need a settled, production-ready library today, its v0.1 status, rough docs, and known mobile issues make ProseMirror or Lexical the safer pick for now.
Same author, new library, no compatibility. Wordgard replaces ProseMirror's "steps" with a simpler delta-style change model, makes node and mark types reusable across schemas, adopts CodeMirror 6's facet-based extensions, relaxes content constraints with programmatic corrections, and handles selection itself via beforeinput. ProseMirror remains supported; Haverbeke has said it is not being abandoned.
Yes. Wordgard is released under the MIT license and free to use, on what its author calls an "abundance" funding model that prioritizes wide adoption. The real cost is the engineering time to build and maintain an editor on top of it.
No. Wordgard is an editor library — it provides the editing surface and document model, and its own site advertises "0% AI." It generates no text, images, or video and publishes nothing. To turn a draft into finished posts across platforms you would use a content engine like Kompozy.
Not yet, by its author's own account. Version 0.1 is described as unproven, documentation is rough, mobile-browser issues have already surfaced, and he expects to stay on 0.x releases for at least a year. It is one to prototype with and watch rather than to build a production editor on today.
Developers building an editing surface should use Wordgard (or ProseMirror). Creators and founders who want to write and then publish across platforms should use Kompozy — it is a content generation and publishing engine, not an editor library. They sit at different points in the workflow and can even be used in sequence.