WordCraft reimagines Microsoft Word as a clean-room open-source app in pure Rust

A Word clone in pure Rust

A project called WordCraft has surfaced on the Hacker News front page, presenting itself as a clean-room reimplementation of Microsoft Word written entirely in Rust. Released by a team called ArtCraft under a dual MIT OR Apache-2.0 licence, the application attempts to reproduce Word's workflow — the ribbon, styles, tables, track changes, references and mail merge — while running natively on macOS, Windows, Linux and BSD, and in the browser via WebAssembly.

According to the project's GitHub repository, the clean-room claim means the implementation was derived from public specifications and observation rather than from Microsoft's code, with every bundled asset either original or openly licensed. The interface uses egui rendering on the GPU, and the repository is explicit that there is no Electron and no Tauri in the stack — one Rust codebase serves every platform.

Feature coverage

The feature list is unusually broad for a young project. Writing features include IME support, smart quotes, automatic corrections, regex find and replace and full undo/redo. Formatting spans character and paragraph options — fonts, spacing, indents, tabs with leaders, borders and shading — alongside a live styles gallery and style sets.

Tables support merged and split cells, banded-row styles, formulas, sorting and rows that break across pages. Page layout handles columns, section breaks, headers and footers with first-page and odd/even variants, watermarks, drop caps and automatic hyphenation. The review toolset includes track changes with accept and reject, comment balloons in the margin, a thesaurus, document comparison, an accessibility checker and a document inspector. References cover tables of contents, footnotes, endnotes, captions, cross-references, an index and citations formatted in APA, MLA, Chicago or IEEE style. Mail merge works from CSV files, and the interface ships in eleven languages including English, Chinese, Japanese, Spanish and Turkish.

On files, WordCraft reads and writes .docx (OOXML) documents that the project says open in Word, imports legacy Word 97-2003 .doc files, and also handles .odt, .rtf, .html, Markdown, LaTeX and plain text. PDF export produces selectable text with links and bookmarks.

Built for agents and the command line

The most distinctive design choice is a uniform command layer: every editor action is a command with an identifier, and the repository counts 422 of them. The same commands drive the ribbon, keyboard shortcuts, an in-app command search, a command-line tool, a JSON control channel and an MCP server.

The CLI can convert documents between formats, dump document structure as JSON, or script edits — the quick-start example applies a text selection and a bold command to a template, then saves the result. The MCP integration is aimed squarely at AI agents: one server handles headless document work, while a second can attach to a running instance, giving an agent tools to type, click, screenshot and verify edits through a document inspector without needing to see the screen. Macros can record and replay sequences of commands.

Performance claims include cached paragraph layout that re-lays out a 188-page document in roughly 1.4 milliseconds after a keystroke, with pages rendered on demand. Spelling and grammar checking run fully offline.

Why it matters

Word's file formats and interaction model remain the de facto standard for document work, and the absence of a credible open implementation has kept users tied to Microsoft's ecosystem or reliant on partial compatibility elsewhere. A clean-room implementation in a memory-safe language, one that treats .docx as a first-class format and runs from a single codebase everywhere from BSD to the browser, is a genuine novelty in this category.

The agent-first architecture may prove the more forward-looking part. By making every editor action scriptable and inspectable, WordCraft treats document editing as a surface AI agents can operate reliably, at a point when commercial suites are adding such capability only gradually. The project also publishes a roadmap, a parity assessment and a generated feature checklist for anyone wanting a detailed measure of how close it currently sits to Word.