Excel MCP Server 🔌 MCP server Open source

What it does

Excel MCP Server lets an AI agent create, read and edit Excel workbooks without Microsoft Excel installed. It is a Python server that works directly on .xlsx files, so an agent can generate reports, fill in data, add formulas and charts, or restructure sheets as part of a larger task. It is a popular choice for spreadsheet work and is useful to developers who need test fixtures, data exports or quick analysis files without writing openpyxl code themselves.

Tools it gives your agent

• Workbooks and sheets: create workbooks and worksheets, read metadata, copy, rename and delete sheets
• Data: write data to ranges and read it back
• Formatting: format ranges, merge and unmerge cells, and list merged cells
• Formulas: apply formulas and validate their syntax
• Analysis: create charts, pivot tables and styled Excel tables
• Ranges, rows and columns: copy and delete ranges, validate ranges, read data-validation rules, and insert or delete rows and columns

Works with

Any MCP client. It supports stdio for local use, Streamable HTTP for remote use and the older SSE transport, which is marked deprecated. The README includes a one-click Cursor install link and a standard mcpServers JSON snippet that suits Claude Code, VS Code and similar clients.

How to install

It runs through uvx, so you need uv installed:

uvx excel-mcp-server stdio

For a remote server, run uvx excel-mcp-server streamable-http and set EXCEL_FILES_PATH to the folder it may use.

Maintenance and safety

The server is MIT-licensed and published on PyPI. Its last commit was in spring 2026, so updates are less frequent than before. With stdio the client passes a file path on each call, so the agent can write anywhere your user account can. With the HTTP transports, paths must be relative to EXCEL_FILES_PATH and traversal is rejected, which gives a safer boundary. It edits files in place, so keep copies of important workbooks.

Who should use it

Developers and analysts who want an agent to produce or modify real Excel files, such as reports, fixtures or exports, from a coding session. If you only need CSV, a plain script is simpler.