OfficeCLI is an open-source command-line tool designed specifically to enable AI agents to programmatically read, edit, and create Microsoft Office documents without requiring Office installation. Written in C#, it provides a single binary distribution that works across platforms and handles Word (.docx), Excel (.xlsx), and PowerPoint (.pptx) files through a unified interface.
The core innovation of OfficeCLI is its built-in HTML rendering engine that converts Office documents to HTML or PNG with high fidelity, allowing AI agents to visually understand document layouts and content. This rendering capability closes what the project describes as the render-look-fix loop, enabling agents to see what they have created and make corrections accordingly. The tool requires no external dependencies and no Office installation, making it lightweight and universally deployable.
For AI agent integration, OfficeCLI provides a skill file that agents like Claude Code, Cursor, Windsurf, and GitHub Copilot can automatically install and use. Agents can manipulate Office documents through simple command-line operations rather than complex programmatic APIs. The project demonstrates this capability through AionUi, a companion desktop application that lets users create and edit Office documents through natural language descriptions, with OfficeCLI handling the underlying document operations.
Word document support includes comprehensive formatting capabilities such as internationalization and right-to-left language support with per-script font slots and BCP-47 language tags, paragraph and run-level formatting, tables with virtual column operations, styles, textboxes and shapes with rotation and gradient effects, headers and footers, image embedding, LaTeX-based equations, mermaid diagrams converted to native editable shapes, comments, footnotes, watermarks, bookmarks, tables of contents, charts, hyperlinks, sections, form fields, content controls, field codes with 22 zero-parameter types, OLE objects, and tracked changes with accept/reject functionality.
Excel support encompasses cell operations including phonetic guides and furigana, formulas with over 350 built-in functions and dynamic array spilling, sheet management with visibility controls and print settings, boolean selectors for filtering rows and columns, tables, multi-key sorting, conditional formatting, charts including box-whisker and pareto types with auto-sorting, pivot tables with multi-field support and persistent filters, slicers, named ranges, data validation, image embedding, sparklines, comments, autofilter, shapes, OLE objects, and CSV/TSV import capabilities.
PowerPoint support enables creation and modification of presentations with full design control, as demonstrated by the repository's showcase of presentations created entirely by AI agents without templates or manual editing.
The project maintains multilingual documentation in English, Chinese, Japanese, and Korean, and operates an active Discord community. The tool is positioned as eliminating the need for 50 lines of Python code and multiple separate libraries, replacing complex document manipulation workflows with single-command operations that AI agents can execute autonomously.
