Create & Edit Excel Spreadsheets with AI on Mac (Claude, Cursor, VS Code) — No Microsoft 365

To let Claude, ChatGPT, Cursor or VS Code create, read, and edit Excel spreadsheets on a Mac, install LMCP — a free, local MCP server for macOS. It reads and writes .xlsx files directly through the file format itself, so your AI can build a new spreadsheet, read data out of an existing one, and update specific cells without a Microsoft 365 subscription, a Graph API app, or Excel even being installed. The same install also creates and reads Word documents and PowerPoint decks, and reads PDFs. Setup takes about two minutes, needs no API keys or OAuth, and every file is written straight to your filesystem — ready to open in Excel, Numbers, or Google Sheets. Nothing goes through a cloud upload step.

Why Give Your AI Access to Excel?

Getting an AI assistant to produce a real spreadsheet — not a wall of markdown you reformat by hand — usually runs into one of a few walls. It prints a table and leaves the copy-pasting to you. Or the “proper” integration turns out to need an Azure app registration, Graph API OAuth scopes, and an active Microsoft 365 subscription just to write one cell. Or a web tool offers to build the file for you, which means your numbers are processed on somebody else's server before you see them.

None of that matches what most people actually want: describe the spreadsheet in plain English and get a finished .xlsx file on the Desktop. With LMCP, that is the whole interaction. Ask your AI to build a budget, read a report and tell you what changed, or turn this week's numbers into a client-ready workbook — it works with the file directly, on your Mac, and hands you back something you can open immediately.

Because the same LMCP install also reads your Mail, Calendar, and OneDrive, the AI is not limited to data you paste in. It can pull the numbers from your own apps first, then write them into the spreadsheet in the same request — the workflows further down show what that looks like in practice.

What You Will Need

• macOS 12 (Monterey) or later — Apple Silicon or Intel
• Any MCP-compatible AI client: Claude Desktop, Cursor, VS Code (Copilot), Windsurf, or ChatGPT
• Nothing else — no Excel installation, no Microsoft 365 subscription, no Office license

You do not need API keys, OAuth tokens, or a Microsoft account. LMCP reads and writes the .xlsx file format directly, so it works whether or not Excel is installed on your Mac at all.

Step 1: Install LMCP on Your Mac

Download LMCP and install it:

1. Open the downloaded LMCP-Installer.pkg from your Downloads folder
2. Follow the installer — it installs Local MCP to Applications for you (no dragging)
3. Open Local MCP from your Applications folder — it appears in your menu bar

Takes about 30 seconds. Your AI clients are configured automatically.

LMCP does not need any macOS automation permission to read or write Office files — it works with the file format on disk, not by driving Excel itself. You will only see a permission prompt if you also ask your AI to touch Mail, Calendar, or another app that does require one.

Step 2: Restart Your AI Client

Quit your AI client completely (Cmd+Q, not just closing the window) and reopen it. It needs to restart to discover the new MCP server.

• Claude Desktop — look for the hammer icon in the chat input
• Cursor — MCP tools appear automatically in the composer
• VS Code — reload the window (Cmd+Shift+P → “Reload Window”)
• ChatGPT — add https://local-mcp.com/mcp as a custom connector

Step 3: Try It Out

Start a new conversation and try these prompts:

• “Create an Excel file with my Q3 expenses: rent 2000, salaries 12000, software 800” — builds a real .xlsx you can open in Excel, with headers and rows already in place.
• “Read budget.xlsx and tell me which months went over plan” — reads a sheet and answers in plain English instead of you scrolling columns.
• “Update cell B5 in forecast.xlsx to 15000” — writes one cell without touching the rest of the file.
• “Find the spreadsheet I downloaded yesterday and summarize the totals” — locates the file on your Mac first, then reads it.
• “Turn these meeting notes into a Word document and save it to my Desktop” — creates a formatted .docx from plain text.

The result lands on your filesystem as a normal file — there is nothing to export or download from a chat window.

What Your AI Can Do with Office Files

LMCP exposes a focused set of tools across the main Office formats:

• Excel — excel_create builds a new .xlsx with headers and data rows from a plain-English description; excel_read pulls data back out of any sheet in an existing workbook; excel_write_cell updates one cell at a time without disturbing the rest of the file.
• Word — word_create generates a new .docx, word_read extracts its text, and word_append adds content to a document that already exists.
• PowerPoint — ppt_create generates a .pptx deck from structured content, and ppt_read reads the text back out of an existing one.
• PDF — pdf_read extracts text from any PDF for summarizing, searching, or pulling numbers into a spreadsheet.

Two things make this more than a single-file trick. First, file_search and finder_search let your AI locate the workbook you mean (“the spreadsheet in Downloads”) before it reads or writes it, so you rarely need to type a full path. Second, the same conversation can reach the other domains LMCP exposes — Mail, Calendar, OneDrive — so the AI can pull source numbers from one app and land them in a spreadsheet in a single request, instead of you copying data between windows yourself.

Limitations

Because LMCP works with the file formats directly, keep these in mind:

• Content, not pixel-perfect styling — LMCP writes data, text, and structure; complex formatting like charts, conditional formatting, or macros is best added in Excel itself afterward.
• Values, not live formulas — excel_write_cell stores whatever value you give it; if you need a working formula in that cell, ask for it explicitly or finish the calculation in Excel.
• One cell per call — updating a whole column means your AI calls excel_write_cell once per cell. Fine for a handful of edits; for rebuilding a large sheet, ask for a new file with excel_create instead.
• Local files — the file lives on your Mac; combine with LMCP's OneDrive tools to read from or save into your synced OneDrive folder.
• Large workbooks — very large spreadsheets read fastest when you tell the AI which sheet you actually need.

Privacy: Everything Stays on Your Mac

This is the key difference between LMCP and cloud-based spreadsheet tools:

• LMCP runs as a local server on your Mac and reads or writes the .xlsx/.docx/.pptx file directly on disk
• There is no upload step — the file never travels to an external server to be built or analyzed
• Results go straight from your filesystem into the AI conversation you are already having
• No spreadsheet data or file content is stored, cached, or logged by LMCP

No API keys, no OAuth flows, no cloud service that could be breached. Your spreadsheets stay exactly where you put them. GDPR and CCPA compliant by architecture — see how that works in more detail in the GDPR-compliant AI guide.

LMCP vs. Editing Excel with Python vs. Uploading the File to Chat

There are two other common ways to get an AI to touch a spreadsheet, and each is a reasonable choice for the right job. Here is the honest comparison:

In short: reach for Python and a library like openpyxl when you need formulas, charts, or formatting that goes beyond values in cells — that is real programmatic control LMCP does not attempt to replace. Upload the file to a web AI for a quick one-off question when you do not mind the data leaving your Mac. Pick LMCP when you want a spreadsheet built or updated in plain English, in two minutes, without writing code, and without your numbers leaving your machine.

Troubleshooting

The AI says it created the file, but I cannot find it

Ask your AI where it saved the file — LMCP writes to the path you gave it, or a sensible default like your home folder if you did not specify one. You can also ask “find the Excel file you just created” and let file_search or finder_search locate it.

Reading a spreadsheet returns nothing or the wrong data

Check that the file path is correct and that the workbook has more than one sheet — if your data lives on a sheet other than the first one, tell your AI which sheet to read by name. Very large workbooks also read more reliably when you point the AI at a specific sheet instead of asking it to scan the whole file.

MCP tools not visible in AI client

Make sure you fully quit and restarted your AI client. You can also check the config file:

• Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
• Cursor: ~/.cursor/mcp.json
• Windsurf: ~/.codeium/windsurf/mcp_config.json

The file opens but formatting looks plain

That is expected: LMCP writes data, text, and structure, not styling. Open the file in Excel afterward for colors, conditional formatting, charts, or anything more than the default look.

Real-World Workflows

The value is not one file — it is having your AI pull data from your other tools and turn it into a finished document.

From data to spreadsheet in one step

Your AI gathers the numbers from Mail and hands you a real .xlsx — no manual entry.

Clean up a messy export

Your AI reads the messy file with excel_read, works out which rows and values need fixing, and writes a new, usable workbook with excel_create instead of you scrubbing it by hand.

Answer questions about a spreadsheet instantly

No pivot table required — your AI reads the sheet and answers directly, which is often faster than opening Excel for a single question.

Turn an email thread into a deck

Because LMCP also reads your email, the AI pulls the source material and produces the deck in one pass.

Report straight into OneDrive

With LMCP's OneDrive tools, the finished file syncs to all your devices automatically.

What Else Can LMCP Do?

Office files are just one domain. LMCP also connects your AI assistant to email, OneDrive, Microsoft Teams, Calendar, Contacts, and local files. Check the full list of guides or visit local-mcp.com to learn more about all 188+ tools available.

Frequently Asked Questions

Do I need Microsoft 365 or Excel installed to use this?

No. LMCP reads and writes the .xlsx file format directly, so it works whether or not Excel is installed and without any Microsoft 365 subscription.

Can my AI edit an existing spreadsheet, or only create new ones?

Both. excel_create builds a new file from scratch, excel_read pulls data out of one that already exists, and excel_write_cell updates individual cells in it without touching the rest of the workbook.

Will the formatting, formulas, or charts come through?

LMCP writes data, text, and structure — not formulas, charts, or conditional formatting. Cell writes store values. For anything beyond that, finish the file in Excel, or use a code-based approach like Python with openpyxl for full programmatic control.

Does my spreadsheet data get uploaded anywhere?

No. LMCP reads and writes the file on your Mac's disk. There is no upload step, and no spreadsheet content is stored, cached, or logged by LMCP.

Does this work with Word and PowerPoint too?

Yes. The same install lets your AI create and read .docx Word documents, create and read .pptx PowerPoint decks, and extract text from PDFs — all with the same local, no-subscription approach as Excel.

Should I use LMCP or write a Python script with openpyxl?

If you need formulas, charts, or precise formatting, a Python script with a library like openpyxl gives you full control and LMCP does not try to replace that. If you want a spreadsheet built or edited in plain English in about two minutes, without writing or running any code, choose LMCP.

Related Guides

• Create Excel Files with AI on Mac (excel_create deep dive)
• Read PowerPoint Slide Text with AI on Mac
• Read & Write OneDrive Files from Claude on Mac — No Graph API
• GDPR-Compliant AI for Mac — How LMCP Keeps Your Data Local