LibreOffice MCP Tools
Enables AI agents to read, write, and edit Office documents via LibreOffice with token-efficient design. Supports multiple formats including DOCX, XLSX, PPTX, and legacy formats through LibreOffice bridge.
README
LibreOffice MCP Tools
[!WARNING] This project was written by GitHub Copilot and has not been fully reviewed by a human. Code may contain bugs, security issues, or unexpected behavior. Use at your own risk. Do not use in production without thorough review.
A Model Context Protocol (MCP) server that gives AI agents (Claude, Copilot, Gemini, Cursor, etc.) the ability to read, write, and edit Office documents via LibreOffice — with a token-efficient design that minimizes LLM context usage.
Inspired by the architecture of chrome-devtools-mcp.
✨ Features
- 22 MCP tools covering reading, writing, spreadsheets, and presentations
- Token-efficient design: outline-first navigation, range-based access, pagination
- Broad format support: DOCX, DOC, XLSX, XLS, PPTX, PPT, ODT, ODS, ODP, RTF, CSV, TXT, PDF
- Legacy format bridge:
.doc,.xls,.pptauto-converted via LibreOffice before parsing - No LibreOffice required for basic reads: native parsers handle DOCX, XLSX, PPTX directly
- LibreOffice required for: legacy formats, PDF export, format conversion
📋 Supported Formats
| Format | Extensions | Read | Write | Method |
|---|---|---|---|---|
| Word 2007+ | .docx, .dotx |
✅ | ✅ | Native (mammoth read / JSZip OOXML write) |
| Word 97-2003 | .doc, .dot |
✅ | ✅ | LibreOffice bridge |
| Excel 2007+ | .xlsx, .xlsm |
✅ | ✅ | Native (ExcelJS) |
| Excel 97-2003 | .xls |
✅ | ✅ | LibreOffice bridge |
| PowerPoint 2007+ | .pptx |
✅ | ✅ | Native (JSZip OOXML) |
| PowerPoint 97-2003 | .ppt |
✅ | ✅ | LibreOffice bridge |
| OpenDocument Text | .odt |
✅ | ✅ | LibreOffice bridge |
| OpenDocument Spreadsheet | .ods |
✅ | ✅ | LibreOffice bridge |
| OpenDocument Presentation | .odp |
✅ | ✅ | LibreOffice bridge |
| Rich Text Format | .rtf |
✅ | ✅ | LibreOffice bridge |
| CSV | .csv |
✅ | ✅ | Native |
.pdf |
✅ (text) | ❌ | LibreOffice CLI | |
| Plain text | .txt |
✅ | ✅ | Native |
🚀 Quick Start
Prerequisites
- Node.js 20+
- LibreOffice (optional for basic DOCX/XLSX/PPTX reads; required for .doc/.xls/.ppt and format conversion)
- Windows: Download LibreOffice
- macOS:
brew install --cask libreoffice - Linux:
sudo apt install libreofficeorsudo dnf install libreoffice
Installation
Using npx (recommended — no install needed):
{
"mcpServers": {
"libreoffice": {
"command": "npx",
"args": ["-y", "@passerbyflutter/libreoffice-mcp-tools"]
}
}
}
Global install:
npm install -g @passerbyflutter/libreoffice-mcp-tools
From source:
git clone https://github.com/passerbyflutter/libreoffice-mcp-tools
cd libreoffice-mcp-tools
npm install
npm run build
Configure your MCP client
Add to your MCP client configuration (e.g., Claude Desktop claude_desktop_config.json):
{
"mcpServers": {
"libreoffice": {
"command": "npx",
"args": ["-y", "@passerbyflutter/libreoffice-mcp-tools"],
"env": {
"SOFFICE_PATH": "/path/to/soffice"
}
}
}
}
Or use .mcp.json at your project root:
{
"mcpServers": {
"libreoffice": {
"command": "npx",
"args": ["-y", "@passerbyflutter/libreoffice-mcp-tools"]
}
}
}
CLI Options
node build/bin/libreoffice-mcp.js [options]
--libreoffice-path <path> Path to soffice executable
(default: auto-detected or SOFFICE_PATH env)
🛠 Tool Reference
Document Management
| Tool | Description |
|---|---|
document_open |
Open a file → returns docId handle. Auto-bridges legacy formats. |
document_close |
Release document handle and temp files |
document_list |
List all open documents |
document_create |
Create new empty document (writer/calc/impress) |
document_save |
Save to current or new path |
document_export |
Export via LibreOffice (PDF, HTML, CSV, etc.) |
document_convert |
Convert file format (DOC→DOCX, XLSX→CSV, etc.) |
Reading (Token-Efficient)
| Tool | Description |
|---|---|
document_get_metadata |
Title, author, word/page count, dates |
document_get_outline |
Headings (Writer) / sheet names (Calc) / slide titles (Impress) |
document_read_text |
Paginated document text as Markdown |
document_read_range |
Specific paragraph or slide range |
document_search |
Find text with surrounding context |
Writing (Writer)
| Tool | Description |
|---|---|
document_insert_text |
Insert at start/end/after heading |
document_replace_text |
Find & replace (first or all occurrences) |
document_insert_paragraph |
Insert paragraph at specific index |
document_apply_style |
Apply heading/paragraph style |
Spreadsheet (Calc)
| Tool | Description |
|---|---|
spreadsheet_list_sheets |
Sheet names with row/col counts |
spreadsheet_get_range |
Cell range as JSON + markdown table |
spreadsheet_set_cell |
Set cell value or formula |
spreadsheet_set_range |
Set 2D range of values |
spreadsheet_add_sheet |
Add new sheet |
spreadsheet_get_formulas |
Get formula expressions in range |
Presentation (Impress)
| Tool | Description |
|---|---|
presentation_list_slides |
Slide titles with index |
presentation_get_slide |
Full slide content (title, body, notes) |
presentation_get_notes |
Speaker notes |
presentation_add_slide |
Add new slide (requires LibreOffice) |
presentation_update_slide |
Update slide content |
💡 Token-Saving Workflow
For maximum token efficiency, follow this pattern:
1. document_open(filePath) → get docId
2. document_get_metadata(docId) → understand size/type
3. document_get_outline(docId) → see structure
4. document_read_range(docId, startIndex=N, endIndex=M) → read specific section
Instead of dumping the entire document, you navigate to exactly what you need.
Spreadsheet workflow:
1. document_open(path) → docId
2. spreadsheet_list_sheets(docId) → see all sheets
3. spreadsheet_get_range(docId, sheetName="Sales", range="A1:D20") → targeted data
🏗 Architecture
src/
├── index.ts # createMcpServer() — MCP server factory
├── LibreOfficeAdapter.ts # soffice subprocess manager
├── DocumentContext.ts # Open document registry
├── DocumentSession.ts # Per-document state + format bridge
├── McpResponse.ts # Response builder (text/JSON/markdown)
├── Mutex.ts # Serializes LibreOffice subprocess calls
├── parsers/
│ ├── DocxParser.ts # DOCX read → {paragraphs, outline, metadata} (mammoth)
│ ├── DocxOoxmlEditor.ts # DOCX write → direct JSZip OOXML manipulation (format-preserving)
│ ├── XlsxParser.ts # XLSX read/write via ExcelJS
│ ├── PptxParser.ts # PPTX read → {slides[]} (JSZip XML)
│ └── PptxOoxmlEditor.ts # PPTX write → add/update slides, create PPTX (JSZip OOXML)
├── formatters/
│ ├── MarkdownFormatter.ts
│ ├── JsonFormatter.ts
│ └── TableFormatter.ts # Spreadsheet → Markdown table
└── tools/
├── documents.ts # open/close/list/create
├── reader.ts # metadata/outline/read/search
├── writer.ts # insert/replace/style
├── spreadsheet.ts # get/set cells/ranges/sheets
├── presentation.ts # slides/notes
└── converter.ts # save/export/convert
🧪 Testing
# Create sample fixtures
node tests/create-fixtures.mjs
# Run smoke tests
npm test
📝 Environment Variables
| Variable | Description |
|---|---|
SOFFICE_PATH |
Path to LibreOffice soffice executable |
DEBUG |
Set to lo-mcp:* for verbose logging |
📄 License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。