WordCraft Collab 1.0: An Open-Source Word Processor Where Humans and AI Agents Work Together

Hossam Taha — Fri, 30 Oct 2026 07:18:32 +0000

What if multiple people and AI agents could edit the same Word document simultaneously, directly inside your application?

That is what I set out to build with WordCraft Collab.

WordCraft Collab 1.0 is an open-source, Word-style document editor built with Rust and WebAssembly. It brings real-time collaborative editing, comprehensive document capabilities, native Arabic and RTL support, and AI agent participation into one application.

It can run as a desktop application, command-line tool, or browser-based editor. It can also be embedded into existing websites using its companion integration package.

GitHub: https://github.com/hossamTahaHussein/wordcraft-collab

The idea behind the project

Document editing is a fundamental capability in many enterprise applications, content management systems, business platforms, and collaboration tools.

However, building a complete word processor is a significant engineering challenge.

I wanted to explore a different approach: a reusable, full-featured document editor that developers can integrate into their own applications, with collaboration built into the editing experience.

I also wanted AI agents to become actual participants in document workflows rather than simply external chat assistants.

1. Real-time collaborative editing

Multiple users can edit the same DOCX document simultaneously.

Every user's changes are reflected across connected browsers, with a collaboration mechanism designed to keep document replicas consistent.

Key capabilities include:

Live cursors and participant names
Real-time presence indicators
Automatic document saving
Saving and synchronization status
Shared undo and redo, while preserving individual edit ownership
Per-user suggesting mode
Edit, suggest, comment, and view permissions
Version history with preview and restore
Offline protection with read-only fallback and Save My Copy

How collaboration works

Every edit is represented as a WordCraft command.

Each browser applies its own edits immediately. A lightweight relay server orders incoming edits, and each connected replica replays the ordered operations while rebasing its own unconfirmed changes.

This architecture allows multiple users to work concurrently without requiring every keystroke to wait for a server response.

The collaboration implementation includes replica management, rebasing, position mapping, selective undo, and protocol handling.

2. Arabic and RTL support

Arabic support was a major consideration in this project.

Rather than treating RTL as a simple text-direction setting, the implementation covers the editor interface, document structure, formatting, and exported files.

Interface support:

Fully translated Arabic interface
Mirrored ribbon, menus, dialogs, panes, rulers, and status bar
Arabic interface through ?lang=ar or system language detection

Document support:

Right-to-left paragraphs, tables, and sections
Mixed Arabic and English text
Correct text ordering
Kashida justification
Arabic spell checking

File support:

RTL preservation in DOCX, HTML, ODT, and RTF
Searchable and selectable Arabic text in exported PDFs
Noto Sans Arabic font bundled with the application
Ayaspell Arabic dictionary bundled for spelling support

Arabic is not an additional plugin or an external dependency. It is part of the editor itself.

3. AI agents as document collaborators

One of the most interesting capabilities is integrating AI agents directly into collaborative documents.

Using the Model Context Protocol (MCP), an AI agent can join a shared document as a named participant:

wordcraft-cli collab-mcp

Agents can:

Execute document editing commands
Insert and modify content
Suggest changes
Comment on document sections
Reply to comments
Identify other participants
Wait for document changes
Inspect document structure and content

Agents operate in suggesting mode by default, allowing human users to review and accept or reject proposed changes.

This creates a different interaction model: instead of copying content between a chatbot and a document, the AI agent works within the document collaboration workflow.

4. Developer integration

WordCraft Collab is designed to be embedded into other applications.

The companion wordcraft-embed project provides the JavaScript SDK, collaboration relay server, and server-side API.

JavaScript SDK

The SDK provides access to the editor's 649 commands through a unified interface:

run(command, params)

It also exposes operations including:

getText
getDocument(format)
inspect
insertText
select
setReadOnly
change events
selectionchange events
presence events
comment events

Server-side integration

The collaboration relay exposes an HTTP API for:

Retrieving document information
Exporting documents as DOCX, PDF, or text
Sending commands as live document edits
Removing participants
Closing documents

It also supports HMAC-signed webhooks for events including:

joined
left
saved
edited
commented
restored

This allows host applications to integrate document collaboration into their existing business workflows.

5. A complete word-processing experience

WordCraft Collab is not just a collaborative text editor.

It includes a broad range of document creation, formatting, layout, review, and publishing capabilities.

Writing and formatting

Smart quotes and AutoCorrect
Lists and automatic list formatting
Clipboard operations with formatting
Find and replace with regular expressions
Font families, sizes, colors, and character spacing
Bold, italic, underline, strikethrough
Superscript and subscript
Format Painter and Change Case

Paragraphs and styles

Alignment and indentation
Line and paragraph spacing
Tabs, leaders, borders, and shading
Built-in styles and custom styles
Style galleries and themes
Widow and orphan control

Tables and pages

Table insertion, merging, and splitting
Table styles and repeated header rows
Table sorting and formulas
Page margins, orientation, and sizes
Multiple columns and section breaks
Headers, footers, and page numbers
Watermarks and page borders

Images and objects

Image insertion, cropping, resizing, and rotation
Brightness and contrast adjustments
Transparency and background removal
Shapes and text boxes
Floating objects and text wrapping

References and review

Tables of contents
Footnotes and endnotes
Citations and bibliographies
APA, MLA, Chicago, and IEEE styles
Captions and cross-references
Index and table of authorities
Spelling and grammar suggestions
Comments and track changes
Accept and reject changes
Document comparison
Accessibility checking and document inspection

Mailings

Mail merge from CSV
Merge fields
Address blocks and greeting lines
Preview and document generation
Envelopes and labels

Viewing

Print layout, web layout, and draft mode
Read mode and focus mode
Zoom and multiple-page views
Navigation pane and rulers
Gridlines and dark mode

6. Multiple document formats

The editor supports reading and writing DOCX files, including importing legacy Word 97–2003 DOC files.

Other supported formats include:

PDF export
ODT
RTF
HTML
Markdown
LaTeX
Plain text
Page images

The objective is to make document creation and conversion available through the same underlying document engine.

7. Desktop, browser, and command line

The project is not limited to a browser-based editor.

It provides three application entry points:

Desktop application

A full desktop word processor with a sample document and support for opening existing documents.

Browser application

A WebAssembly-powered editor that runs in the browser and supports real-time collaboration.

Command-line interface

The CLI enables document conversion, text extraction, inspection, and automated document editing.

For example:

wordcraft-cli convert report.docx report.pdf
wordcraft-cli text report.docx
wordcraft-cli inspect report.docx

You can also execute document operations through command sequences or JSON scripts.

8. Architecture and technology

The project is built around a layered Rust architecture.

Layer	Components
L0	Geometry and measurements
L1	Document model, fonts, spelling and grammar
L2	Layout, pagination, DOCX and other formats
L3	Rendering and PDF generation
L4	Editing engine, sessions, undo/redo, 649 commands
L5	MCP server and collaboration engine
L6	UI built with egui
Applications	Desktop, CLI and WebAssembly browser application

The browser application uses Rust compiled to WebAssembly.

The architecture separates document modeling, editing, rendering, collaboration, and user interface responsibilities.

This makes it possible to expose the same underlying document capabilities through different application interfaces.

9. Testing and reliability

Collaborative editing requires more than verifying that two users can type simultaneously.

The project includes:

Randomized multi-user convergence tests
Undo and redo collaboration tests
Unit tests
End-to-end tests with three browser instances
Relay restart testing
AI agent integration testing against the actual collaboration relay

The CI workflow runs formatting checks, Clippy, approximately 800 tests, asset attribution checks, architecture layering checks, and the WebAssembly build.

10. Getting started

Clone the repository:

git clone https://github.com/hossamTahaHussein/wordcraft-collab
cd wordcraft-collab

Run the desktop application:

cargo run --release -p wordcraft -- --sample

Open an existing document:

cargo run --release -p wordcraft -- report.docx

Run the browser version:

cd apps/wordcraft-web
trunk serve

Open http://127.0.0.1:8771/?sample

For the browser collaboration setup and embedding instructions, see the companion repository. A prebuilt web build is attached to the v1.0.0 release.

Open source and contributions

WordCraft Collab is dual-licensed under MIT or Apache-2.0, at your option.

The project is based on WordCraft by the ArtCraft team and its contributors. WordCraft Collab is an independent modified project and is not sponsored, affiliated with, or endorsed by the ArtCraft team.

The collaboration engine, Arabic and RTL improvements, integrations, and AI agent capabilities are part of WordCraft Collab.

Explore the repositories:

Editor: https://github.com/hossamTahaHussein/wordcraft-collab
Embedding SDK and collaboration relay: https://github.com/hossamTahaHussein/wordcraft-embed

What's next?

Version 1.0 establishes the foundation for a different approach to document collaboration.

Coming in the next release: spreadsheet and presentation apps. WordCraft Collab will grow beyond documents, with a spreadsheet editor and a presentation editor joining the word processor.

I am particularly interested in expanding the possibilities around AI agents, document automation, and embedding collaborative editing into enterprise applications.

Imagine contracts, reports, proposals, and business documents where people and AI agents can work together, with changes visible, reviewable, and tracked.

That is the direction I want to explore.

I'd like to hear from developers:

What would you build if you could embed a complete word processor into your application and allow AI agents to collaborate directly with your users?

Contributions, feedback, and GitHub stars are welcome.

Website: https://hossamtaha.me

DEV Community: Hossam Taha