Docx-CLI Allows AI Agents to Edit Word Docs Faster and More Accurately

Introduction to Docx-CLI

Docx-CLI is a command-line interface designed to facilitate AI agents in reading and editing .docx files. It allows users to leave comments and suggest inline edits that can later be accepted or rejected within Microsoft Word.

How It Works

Unlike previous methods where agents unzip and manually edit OOXML, Docx-CLI uses stable locators with character offsets. This preserves the original document's formatting and embedded content during the editing process.

The tool mutates XML in place, ensuring that custom styles and theme colors remain intact. It is designed for simplicity and minimizes the need for complex reasoning by the AI.

Performance Comparison

In a controlled A/B testing scenario involving six typical document tasks, Docx-CLI consistently outperformed the default editing skill. Results showed that using traditional means, AI models struggled with correctness, failing to produce openable documents in numerous cases.

The tests revealed that the default method not only incurred higher token consumption (2.2–2.6 times more) but also took longer to complete tasks (1.7–2 times slower) compared to Docx-CLI.

Installation and Use

Docx-CLI can be installed via npm or through standalone binaries, supporting multiple operating systems. Users can install it easily and run commands to read and edit documents as required. It includes built-in verification for downloaded binaries to ensure integrity.

Conclusion

Docx-CLI addresses a significant need in the realm of document editing for AI agents, showcasing improved efficiency and accuracy. This advancement not only streamlines workflows but also reduces the likelihood of compatibility issues with Microsoft Word, marking a notable step forward in the integration of AI in document handling.