Introducing Excel MCP Server

Your AI.
Real Excel.

Ask in plain English.

Connect your AI assistant to
Microsoft Excel on your desktop.

Real Excel · synthetic sales data
Actual Excel sales table with twelve synthetic orders
Claude, Copilot, and other MCP-compatible assistants
01 / 11
Understand the difference

Same workbook format.
Different engines.

File-based workflows

Anthropic's spreadsheet skill

openpyxl edits the file.
LibreOffice calculates formulas.

openpyxl-based MCP servers

Read and write workbook files.

Excel MCP Server

The real Excel application

Excel opens, calculates,
refreshes, and saves the workbook.

Power QueryDAXVBA
Not just a different interface. A different engine doing the work.
02 / 11
Choose it for the work you actually do

Do you need
real Excel?

Refreshable business reports
Existing models, macros, and PivotTables
Live results from Excel's own engine
A file-based tool may be enough

Simple exports.
No desktop Excel.

Choose a file-based workflow when you need portability, or just a new spreadsheet.

Excel MCP requires Windows + desktop Microsoft Excel.
03 / 11
01 / Start with a question

One workbook.
Two ways to work.

MCP / your AI assistant

“Turn these sales orders
into a regional report.”

Conversational workflow
excelcli / your terminal

The same Excel operations.
Scripted. Repeatable.

Command-line workflow
sales-workbook.xlsx → real desktop Excel
04 / 11
02 / Power Query

Refreshable.
Not disposable.

From raw orders
to analysis-ready data.

1 Read the source
2 Set column types
3 Clean the regions
Actual Excel output / CleanSales
Power Query output with typed sales data
M / excerpt from executed query
Table.TransformColumnTypes(Source,
  {{"Revenue", Currency.Type}})
05 / 11
03 / Data Model + DAX

One definition.
Every decision.

Persisted measure / CleanSales
Total Revenue :=
  SUM(CleanSales[Revenue])
Sales→CleanSales→Data Model
North$194k
South$165k
West$225k
Verified DAX query result · 12 synthetic orders
06 / 11
04 / PivotTables + charts

See the story. Keep the objects.

Actual Excel capture / native PivotTable + linked PivotChart
Actual Excel sales report with native PivotTable and green PivotChart
$225kWest leads revenue
07 / 11
05 / VBA

Keep your
proven
workflows.

Read, edit, and run VBA.
Only run macros you trust.

Illustrative VBA / not executed
Sub RefreshReport()
  Worksheets("Report") _
    .PivotTables("SalesByRegion") _
    .RefreshTable
  Application.Calculate
End Sub
Editing VBA projects needs manual trust access.
08 / 11
06 / Python in Excel

Analysis, right
beside your data.

Genuine repository screenshot / existing product example
Real Excel formula bar showing Python reading a sales revenue column
Microsoft-hosted Python

Supported Microsoft 365.

Python enabled.
Internet required.

09 / 11
07 / Real calculation

Change an input.
Excel does the math.

Report!B3 / verified in Excel
=SUM(Sales[Revenue])
Original total$584k
→
After +$10k to one order$594k
Formulas · Queries · Charts · Workbook features
10 / 11
Choose your way in

Two equal ways in.
One real Excel engine.

MCPConversational assistants
excelcliCan use fewer tokens for coding agents
excelmcpserver.dev
Windows + desktop Excel required
11 / 11
Excel MCP Server lets your AI assistant work in the real Excel desktop app.
Ask in plain English. Excel does the work.
File-based tools are useful for creating and editing spreadsheets without Excel.
Anthropic's spreadsheet skill uses openpyxl, with LibreOffice for calculation.
Excel MCP uses Microsoft's Excel engine instead, including the features built into the app.
Use it for refreshable reports, complex existing workbooks, and live Excel results.
For simple exports, or work without Windows and Excel, file-based tools may be the better fit.
Start with one sales workbook.
Ask your assistant to turn raw orders into a report.
Or drive the same workflow with excelcli.
Power Query turns preparation into a repeatable process.
Clean the source, shape your sales data, and refresh the query inside Excel.
Not a static export.
Bring those tables into the Data Model.
Define revenue once with DAX, then reuse that business logic across your analysis.
The model stays in your workbook.
Build a PivotTable to compare regions.
Add a chart, and the story becomes clear.
These are native Excel objects, ready for the next question.
Already have a trusted workflow?
Work with VBA modules and macros.
Run only trusted macros. They have your user permissions.
Go further with Python in Excel.
Write analysis beside your data.
Python runs in Microsoft's cloud, with a supported subscription and internet access.
Change an input. Excel calculates the result.
Keep the formulas, queries, charts, and workbook features that make the file valuable.
For coding agents, excelcli can use fewer tokens.
Two equal ways in. One real Excel engine.