Excel MCP Server is a Model Context Protocol server that lets AI assistants read and write Microsoft Excel workbooks directly. Instead of exporting spreadsheets to CSV and losing structure, an MCP-compatible client can open .xlsx files, inspect values and formulas, and write results back into the sheet. The project is MIT licensed and has earned over 1,000 stars on GitHub.
The server exposes tools for reading and writing cell text and formulas and for creating new sheets, covering the common workbook formats xlsx, xlsm, xltx, and xltm. On Windows it goes further, with live editing of workbooks you already have open and the ability to capture a screenshot image of a sheet — handy when an agent needs to verify a layout visually before declaring a task done. Installation is a single MCP configuration block: the server is distributed through npm and runs on Node.js 20 or later, so adding real spreadsheet access to a desktop AI assistant or coding agent takes minutes rather than a custom integration project.
For AI agents, direct workbook access changes what is feasible: instead of asking a user to copy-paste table contents into a chat, the assistant can inspect a budget sheet itself, add formulas that stay live in Excel, and create a summary sheet in the same file. Keeping formulas as formulas — rather than flattened values — means the workbook remains fully editable and auditable in Excel afterwards, which matters for finance and operations work where a spreadsheet is a living document rather than a static export.
Excel MCP Server is free and open source under the MIT license.
It is for analysts, accountants, and operations teams who live in Excel and want their AI assistant to work inside real workbooks instead of pasted CSV dumps.