OfficeCLI: AI Agent Office Suite CLI

1. Project Overview

OfficeCLI is a single-binary command-line tool that lets AI agents and developers create, read, analyze, and modify Word, Excel, and PowerPoint documents — including high-fidelity visual rendering — without ever installing Microsoft Office.

2. Background & Positioning

OfficeCLI describes itself as "the world's first and the best Office suite designed for AI agents." Its core mission is to give autonomous AI systems a reliable, structured, and inspectable way to produce real Office files, closing the gap between an LLM's text output and a finished .docx/.xlsx/.pptx deliverable a human can open and trust.

Compared to traditional approaches — driving Microsoft Office/LibreOffice via COM/UNO automation, or scripting with Python libraries like python-docx and openpyxl — OfficeCLI is built AI-native from the ground up: it ships as a single self-contained binary with an embedded .NET runtime (no Office installation, no external dependencies), exposes a path-based addressing model (e.g. /slide[1]/shape[2]) that maps cleanly onto how an LLM reasons about documents, and includes a built-in rendering engine so an agent can visually verify its own output before handing it off.

3. Feature Categories

📄 Document Operations — 20 commands (create, view, get, set, add, remove, move, swap, batch, validate) — Full CRUD lifecycle for .docx, .xlsx, and .pptx files through a consistent, path-addressable command set.
🎨 Rendering & Preview — HTML, PNG, PDF, and SVG output — Converts documents into high-fidelity visuals so agents (and humans) can inspect results, including a live-refreshing watch command.
📊 Excel Engine — 350+ built-in functions — Formula evaluation, dynamic arrays, and native pivot table generation directly from data ranges, no spreadsheet engine required.
🔁 Data & Templates — merge, dump — {{key}} placeholder template merging for batch document generation, and round-trip JSON serialization for replayable document blueprints.
🤖 AI Agent Integration — MCP server, auto-installed skills — Registers as a Model Context Protocol server (officecli mcp claude) and auto-detects/installs skill files into tools like Claude Code, Cursor, and GitHub Copilot.

4. Key Highlights

Zero-dependency single binary — Embeds the .NET runtime, so there is no Office installation, no COM automation, and no fragile system dependency chain.
Three-layer architecture — L1 semantic views (text, outline, stats), L2 structured DOM operations (get/set/add/remove via path), and L3 raw XPath access for edge cases — giving agents both high-level convenience and low-level control.
Built-in visual rendering — Documents can be rendered to HTML/PNG/PDF/SVG on demand, letting an AI agent "see" what it produced instead of guessing from raw XML.
Resident mode — officecli open keeps a document in memory across multiple commands for near-zero-latency multi-step editing workflows.
Atomic batch operations — Multiple edits can be grouped into a single transaction via batch, reducing the risk of partially-applied changes.
Native MCP support — Works out of the box with Claude Code, Cursor, VS Code, and other MCP-compatible AI tools, with SDKs for Python and Node.js for deeper programmatic integration.

5. Use Cases by Role

General Developers — Automate report, invoice, or slide-deck generation in CI pipelines or backend services without bundling Office or LibreOffice.
AI Agent Builders — Give coding/office agents a dependable tool to produce and verify real Office deliverables as part of an agentic workflow, via the built-in MCP server.
Data/Research Scientists — Generate Excel workbooks with computed formulas and pivot tables directly from data pipelines, or produce templated Word/PowerPoint reports from structured data.
Project Managers — Use merge with data files to batch-generate status reports, proposals, or presentation decks from a single template.

6. Getting Started

Find what you need — browse the command reference and wiki for format-specific guides:

officecli --help
officecli view <file> outline

Install and integrate:

curl -fsSL https://raw.githubusercontent.com/iOfficeAI/OfficeCLI/main/install.sh | bash
# Windows PowerShell
irm https://raw.githubusercontent.com/iOfficeAI/OfficeCLI/main/install.ps1 | iex
# Register as an MCP server for Claude Code
officecli mcp claude

Contribute:

git clone https://github.com/iOfficeAI/OfficeCLI.git
# Build requires the .NET 10 SDK; output is a self-contained native binary

Open a pull request against the main branch, or join the community Discord to discuss ideas first.

7. Project Structure

OfficeCLI/
├── src/officecli/     # Core CLI source code
├── sdk/                # Python and Node.js SDKs
├── skills/             # AI agent skill definitions (Claude Code, Cursor, etc.)
├── plugins/            # Extensibility framework
├── schemas/            # OpenXML validation schemas
├── examples/           # Runnable example scripts
├── npm/                # Node.js package distribution
├── assets/             # Visual demonstrations and showcases
└── .github/workflows/  # CI/CD automation

src/officecli holds the three-layer document engine; skills/ is what gets auto-installed into detected AI coding tools; sdk/ is the entry point for teams that want programmatic access instead of shelling out to the CLI.

8. Related Ecosystem

Upstream/runtime: .NET 10 SDK (build-time), OpenXML format specifications (the schemas OfficeCLI validates against).
AI tool integrations: Model Context Protocol (MCP), Claude Code, Cursor, VS Code, GitHub Copilot — OfficeCLI plugs into these as a document-manipulation backend.
Complementary/alternative tooling: Traditional automation via Microsoft Office/LibreOffice COM/UNO, and Python libraries such as python-docx and openpyxl — OfficeCLI targets the same problem space with an AI-native, dependency-free CLI instead.

9. License

Licensed under the Apache License 2.0.

✅ Free to use, modify, and distribute, including in commercial and closed-source products.
✅ Includes an express patent grant from contributors.
❌ No warranty is provided; the software is offered "as is."
ℹ️ Modified versions must retain copyright, license, and attribution notices, and clearly state any changes made.

10. FAQ

Q: Do I need Microsoft Office installed to use OfficeCLI?
A: No. OfficeCLI ships as a self-contained binary with an embedded .NET runtime and manipulates OpenXML files directly.

Q: Which file formats are supported?
A: .docx (Word), .xlsx (Excel), and .pptx (PowerPoint), including Excel formula evaluation and pivot tables.

Q: How does OfficeCLI integrate with AI coding assistants?
A: It runs as an MCP server (officecli mcp claude) and can auto-install skill files into supported tools such as Claude Code, Cursor, VS Code, and GitHub Copilot.

Q: Can I preview a document without opening it in Office?
A: Yes — use officecli watch <file> for a live-refreshing browser preview, or officecli view <file> to render HTML/PNG/PDF/SVG output.

Q: Is there a programmatic SDK, or only the CLI?
A: Both — install officecli-sdk (Python) or @officecli/sdk (Node.js) to call OfficeCLI functionality directly from code.

11. Quick Links

Repository: https://github.com/iOfficeAI/OfficeCLI
Official site: https://officecli.ai
Documentation/Wiki: https://github.com/iOfficeAI/OfficeCLI/wiki
Community (Discord): https://discord.gg/2QAwJn7Egx

12. Summary

OfficeCLI fills a specific gap in the AI tooling landscape: a dependency-free, path-addressable, visually verifiable way for agents to produce real Office documents. Developers building document-generation pipelines, and teams building AI agents that need to output trustworthy Word/Excel/PowerPoint files, are the primary audience — the built-in rendering and MCP integration make it especially well suited to agentic workflows where the AI itself needs to check its own work.