TL;DR:
Describe what you need in plain language and let AI generate formulas, clean messy data, build charts, and automate multi-step reporting workflows directly inside Excel.
Works with ChatGPT Work, Claude Cowork, and Cursor via a single Composio MCP connection.
Composio handles OAuth token refresh automatically so connections don't break mid-session.
Setup takes under 30 minutes. Free tier includes 20,000 tool calls per month, no credit card required.
Instead of manually entering data, building formulas, cleaning rows, and creating charts, you can describe what you need in plain language and let the AI handle much of the execution. This makes spreadsheet work faster and more accessible, especially for teams managing sales data, financial reports, customer records, project trackers, or operational dashboards. AI can help structure a workbook, update information, detect inconsistencies, generate formulas, surface insights, and automate multi-step reporting workflows.
This guide shows you how to build the connection between Excel and ChatGPT, Claude Cowork, or Cursor. MCP (Model Context Protocol) is the connection standard Composio uses to plug your AI assistant into outside tools like Excel: think of it as the cable that lets your AI talk to your spreadsheets.
Follow this path if your team works out of ChatGPT Work.
How to connect Excel to ChatGPT Work
Prerequisites
Before connecting Excel to ChatGPT Work using Composio, make sure you have:
A ChatGPT account
Access to the Microsoft Excel workspace you want to connect
A Composio MCP account
Permission to authenticate your Excel account through OAuth (you'll be prompted to approve the required permissions during setup)
Step 1: Open the MCP settings
Open ChatGPT Work → Settings → Plugins → MCP. This is where you'll manage MCP server connections.
Step 2: Add the Composio MCP server
Click Add server → enter Composio as the server name → select Streamable HTTP as the server type → paste the following URL:
https://connect.composio.dev/mcp
Click Save. The Composio server will now appear in your list of MCP plugins.
Step 3: Authenticate your account
Click Authenticate → sign in to your Composio account → approve the requested OAuth permissions by selecting Allow access. Once approved, ChatGPT Work can securely access your connected Excel account.
Step 4: Start using Excel with ChatGPT
Once authentication is complete, Excel tools become available inside ChatGPT Work. You can start issuing natural language commands, such as adding rows to spreadsheets, creating charts, sharing workbooks, or clearing worksheet ranges, without manually interacting with the Excel interface.
If you use the Claude desktop app day-to-day, start here.
How to connect Excel to Claude Cowork
Prerequisites
Before connecting Excel to Claude Cowork using Composio, make sure you have:
A Claude Cowork account with the Claude desktop app installed
A Microsoft Excel account you want to connect
A Composio account
Permission to authenticate your Excel account through OAuth in your browser
Internet access to connect to the Composio MCP server
Step 1: Open the Connectors menu
Open Claude Desktop → Customize → Connectors → click the + icon at the top to add a new connector.
Step 2: Add the Composio MCP server
Click Add custom connector → paste the Composio MCP server URL:
https://connect.composio.dev/mcp
→ Click Connect.
Step 3: Authorize Composio
Click Connect → you'll be redirected to your browser → sign in to Composio if prompted → approve the requested permissions to authorize the connection.
Step 4: Connect your Excel account
Return to Claude Cowork → ask Claude to connect to Excel or give it an Excel-related task → when prompted, authenticate your Microsoft account and approve access.
Once complete, Claude Cowork can use Excel through Composio to perform actions such as adding rows, creating charts, managing worksheets, and sharing workbooks using natural language.
For developers already using Cursor, this section covers adding Excel access to your existing workflow.
How to connect Excel to Cursor
Prerequisites
Before connecting Excel to Cursor using Composio, make sure you have:
Cursor installed on your computer
A Composio account
A Microsoft Excel account
Permission to authenticate your Microsoft account through OAuth
Access to Cursor's MCP configuration (
mcp.json) or the ability to use the one-click installer
Step 1: Install Composio
Choose one of the available installation methods:
Option 1 (Recommended): Click Install in Cursor on the Composio website → Cursor automatically adds the Composio MCP server.
Option 2 (Manual): Open .cursor/mcp.json (or ~/.cursor/mcp.json for a global installation) → add the following configuration:
{
"servers": {
"composio": {
"type": "http",
"url": "https://connect.composio.dev/mcp"
}
}
}→ Save the file.
Step 2: Authorize Composio
Restart Cursor → open MCP Tools → click Connect next to Composio → your browser opens automatically → sign in to your Composio account if prompted → approve the authorization request.
Step 3: Connect Excel
Return to Cursor → ask the AI to perform any Excel-related task, such as creating a workbook or adding a row → Cursor prompts you to connect Excel → sign in to your Microsoft account → approve the requested permissions.
Step 4: Start using Excel with Cursor
Once authentication is complete, Cursor can interact with Excel using natural language. You can create workbooks → add worksheets → insert or update table rows → generate charts → export workbooks to PDF → manage permissions → search and update files without leaving Cursor.
Core capabilities of AI in your spreadsheet
AI changes how you work with spreadsheets by letting you manage data, perform analysis, build reports, and automate multi-step tasks using natural-language instructions.
Instead of manually editing cells, writing complex formulas, creating charts, and checking thousands of rows, you can describe the result you need. The AI can then carry out the relevant spreadsheet actions, provided it has access to the workbook and the required tools.
The following seven capabilities cover the most valuable ways AI can support spreadsheet work.
1. Create and organize workbooks
AI can create a complete workbook from a simple description of what you are trying to manage.
Rather than opening a blank file and deciding how to structure it, you can ask AI to create the workbook, add the required worksheets, name each tab, define column headers, and arrange the file around a specific workflow.
For example, a sales workbook may include separate sheets for leads, opportunities, closed deals, monthly performance, and an executive dashboard. A finance workbook may include revenue, expenses, cash flow, budget variance, and forecasts.
AI can help with tasks such as:
Creating a new workbook
Adding and naming worksheets
Renaming existing tabs
Reordering worksheets
Duplicating template sheets
Deleting unused worksheets
Creating column headers
Structuring tables for consistent data entry
Separating raw data from reporting sheets
This is particularly useful when building recurring reports, operational trackers, project plans, budget models, or client-facing workbooks.
A strong prompt should explain what the workbook is for, which sheets are needed, and what information each sheet should contain.
Example prompt
Create a workbook called "2026 Sales Operations" with five worksheets: Leads, Opportunities, Closed Deals, Monthly KPIs, and Executive Dashboard.
In the Leads worksheet, add columns for Company, Contact Name, Email, Lead Source, Industry, Estimated Value, Owner, Status, and Date Added.
In the Opportunities worksheet, add columns for Company, Deal Stage, Deal Value, Probability, Expected Close Date, Owner, and Next Step.
Format the first row of each worksheet as headers and convert each data range into a structured table.
2. Enter, update, and manage spreadsheet data
AI can add, edit, move, and remove spreadsheet data without requiring you to manually navigate individual cells.
This is one of the most practical capabilities because many spreadsheet workflows involve repetitive data maintenance. You may need to add new customer records, update product prices, change invoice statuses, or remove outdated entries.
Instead of locating every relevant row yourself, you can describe the condition that identifies the records and explain what should change.
AI can help with:
Adding individual rows
Importing batches of records
Updating values that match defined conditions
Replacing text across a worksheet
Filling blank cells
Moving records between worksheets
Deleting outdated records
Removing rows that meet specific criteria
Updating dates, categories, or statuses
Applying changes across multiple worksheets
For example, you could ask AI to mark all unpaid invoices older than 60 days as overdue. You could also ask it to move all completed projects to an archive worksheet.
The main advantage is that you can describe the business rule rather than manually applying it row by row.
However, prompts should be precise. Clearly identify the worksheet, the relevant columns, the matching condition, and the required change.
Example prompt
In the Invoices worksheet, find every row where Payment Status is "Unpaid" and the Due Date is more than 30 days in the past.
Change the Payment Status to "Overdue", add today's date to the Review Date column, and enter "Follow-up required" in the Notes column.
Do not change any rows where the invoice is already marked as Paid, Disputed, or Cancelled.
3. Clean and standardize data
AI can identify and correct inconsistent data that would otherwise make reporting, filtering, and analysis unreliable.
Spreadsheet data often contains formatting problems, especially when information has been copied from several systems or entered manually by different people.
Common problems include:
Duplicate records
Inconsistent capitalization
Extra spaces
Missing values
Different date formats
Incorrect number formats
Inconsistent country names
Multiple spellings of the same category
Phone numbers in different formats
Email addresses with accidental spaces
Numbers stored as text
AI can review the dataset, detect these issues, and apply standardized rules.
For example, it can change values such as “United States,” “USA,” “U.S.,” and “US” into one approved format. It can also identify likely duplicates using a combination of company name, email address, customer ID, or phone number.
This capability is especially valuable before creating dashboards, importing data into a CRM, running financial analysis, or sharing reports with stakeholders.
A good cleaning prompt should explain which fields can be standardized automatically and how uncertain cases should be handled. It is often safer to ask AI to flag ambiguous records instead of deleting or changing them automatically.
Example prompt
Clean the Customer Database worksheet using the following rules:
Remove exact duplicate rows.
Flag possible duplicates where the email address, phone number, or customer ID matches.
Trim unnecessary spaces from all text fields.
Convert names and company names to title case.
Convert all email addresses to lowercase.
Standardize dates as YYYY-MM-DD.
Standardize country values using full country names.
Replace blank values in the Industry column with "Unknown".
Do not delete possible duplicates automatically. Highlight them for review.
4. Generate and apply formulas
AI can create spreadsheet formulas from plain-language instructions, reducing the need to remember exact Excel syntax.
This capability is useful for both simple calculations and more advanced spreadsheet logic.
AI can help generate formulas for:
Totals and averages
Conditional calculations
Lookups
Date calculations
Running totals
Percentage changes
Profit margins
Budget variance
Commission calculations
Risk categories
Multi-condition filtering
Error handling
It can produce formulas using functions such as:
SUMSUMIFSUMIFSCOUNTIFCOUNTIFSAVERAGEIFIFIFSXLOOKUPINDEXMATCHFILTERIFERRORTODAYDATEDIFLET
AI can also apply the formula down an entire column and explain what each part does.
For example, you could ask it to calculate expected revenue by multiplying deal value by close probability. You could ask it to categorize invoices based on the number of overdue days or calculate year-over-year growth.
The most effective prompts explain the column names, calculation logic, and expected output.
Example prompt
In the Opportunities worksheet, create a formula in the Expected Revenue column.
The formula should multiply Deal Value by Probability.
Probability is stored as a percentage. If either Deal Value or Probability is blank, return a blank cell instead of zero or an error.
Apply the formula to every populated row in the table and format the result as currency.
5. Analyze data and answer business questions
AI can examine spreadsheet data and explain what is happening in a way that is easier to understand than reviewing rows manually. It can analyze historical performance, compare categories, identify outliers, calculate changes, and surface patterns that may require attention.
AI can help answer questions such as:
Which products generate the most revenue?
Which regions are underperforming?
Which customers are at risk of churning?
Where are expenses increasing?
Which months show seasonal demand?
Which invoices are creating the largest cash-flow risk?
Which campaigns have the best conversion rate?
Which sales representatives are above or below target?
Common analytical tasks include:
Comparing periods
Calculating growth rates
Ranking products or teams
Identifying the largest increases and decreases
Detecting unusual values
Finding correlations
Summarizing performance by category
Highlighting missing or inconsistent data
Calculating key performance indicators
Identifying likely causes of performance changes
AI is particularly valuable when the user does not know which formula, pivot table, or filter is required. You can ask a business question directly and allow the AI to determine how to inspect the data.
However, AI-generated analysis should still be reviewed. A spreadsheet can show relationships and patterns, but these do not always prove causation.
Example prompt
Analyze the Sales Data worksheet and provide the following:
Total revenue by month
Month-over-month growth
Revenue by region
Revenue by product category
The five highest-value customers
The five largest declines compared with the previous quarter
Any unusual spikes or drops that may require investigation
Summarize the findings in plain English and show the calculations used for each conclusion.
6. Create charts, reports, and dashboards
Instead of manually selecting ranges, choosing chart types, configuring labels, and positioning each visual, you can describe the report you need. AI can create:
Line charts
Bar charts
Column charts
Pie charts
Area charts
Scatter plots
KPI summary tables
Pivot tables
Conditional formatting
Executive dashboards
Monthly reporting sheets
Performance scorecards
AI can also choose a suitable chart type based on the data.
For example:
A line chart is useful for showing changes over time.
A bar chart is useful for comparing categories.
A stacked column chart is useful for showing composition.
A scatter plot is useful for examining relationships between variables.
KPI cards are useful for presenting headline metrics.
A dashboard prompt should identify the intended audience, the metrics that matter, the reporting period, and any required comparisons. A leadership dashboard may prioritize revenue, growth, profitability, pipeline, and risk. An operational dashboard may focus on workload, backlog, response time, or completion rates.
Example prompt
Create an Executive Dashboard worksheet using the data from the Sales Data worksheet.
Include:
Total revenue
Total closed deals
Average deal size
Month-over-month revenue growth
Year-to-date revenue versus target
A line chart showing monthly revenue
A bar chart showing revenue by region
A column chart showing closed deals by sales representative
A table showing the ten largest open opportunities
Place the headline KPIs at the top and the charts below them. Use clear titles and make sure all charts update when new rows are added to the source table.
7. Automate multi-step spreadsheet workflows
Combining several actions into one workflow removes more manual steps than any single capability covered here. Many spreadsheet tasks follow the same sequence:
Import new data.
Clean the data.
Update formulas.
Analyze performance.
Refresh charts.
Create a summary.
Export or share the report.
Without AI, each step may require a different menu, formula, or manual process. With AI, you can describe the full outcome in one prompt. This is especially useful for recurring workflows such as:
Weekly sales reports
Monthly finance reviews
Inventory updates
Marketing performance reports
Customer health checks
Expense reconciliation
Project status reporting
Invoice follow-up
Forecast updates
AI can also apply conditional logic within a workflow.
For example, it can update the report only when new data is available, flag incomplete records, avoid overwriting approved values, and create a separate exception sheet for issues that require human review.
The best workflow prompts define the sequence of actions, the source data, validation rules, and the final output.
Example prompt
Update the monthly sales reporting workbook using the latest data in the Raw Sales Import worksheet.
Complete the following steps:
Remove exact duplicates.
Standardize all date fields as YYYY-MM-DD.
Flag rows with missing Deal Value, Close Date, or Sales Representative.
Add valid new rows to the Sales Data table.
Refresh all formulas and pivot tables.
Update the Executive Dashboard.
Calculate revenue, average deal size, win rate, and month-over-month growth.
Create a short summary of the most important changes.
Add any data-quality issues to a separate worksheet called Review Required.
Do not overwrite manually entered notes or approved forecast values.
Start your first connected workflow with a free Composio account (no credit card required, 20,000 tool calls per month). Your AI assistant will be reading and writing to Excel in under 30 minutes.
FAQs
Is it safe to let an AI assistant access my financial spreadsheets?
Yes, if you use a connection layer that runs code in a sandboxed environment and controls which toolkits the AI can access at the team or user level. Avoid uploading raw financial files directly to public AI tools that process data on shared servers without explicit data retention policies.
Do I need to write code to connect my AI to Excel?
No. You can use Composio's pre-built integrations and MCP server to connect your assistant in under 30 minutes without writing any custom glue code. Paste the Composio MCP endpoint URL into your AI client configuration, authorize the Microsoft connection through the Composio dashboard, and the tools are immediately available in your AI assistant.
What happens when my Excel API connection expires?
Microsoft Graph API access tokens expire on a short cycle. Composio's managed auth layer automatically refreshes tokens in the background before they expire, so your workflows never break mid-session.
Key terms glossary
Managed OAuth: A system where a third-party platform handles the secure storage, authorization, and automatic refresh of your API access tokens, so you never manually re-authenticate mid-workflow.
Tool Router: A feature that automatically directs your AI assistant's requests to the correct application (Excel, Google Sheets, Salesforce, etc.) based on your active connections, removing the need to write custom conditional logic in your agent code.
Sandboxing: A security practice that runs software or code execution in an isolated environment to prevent it from accessing or modifying data outside the defined scope. This is a hard requirement when an AI runs code on financial spreadsheets, since any execution that escapes the sandbox can read or overwrite data it was never meant to touch.
Microsoft Graph API: The unified API endpoint for Microsoft 365 services (Excel, OneDrive, Outlook, Teams) that external AI assistants use to read and write to your files with your authorized permission.
RBAC (Role-Based Access Control): A permission system that controls which toolkits or integrations are available to specific users or teams, limiting which connected applications an AI assistant can access rather than individual actions within them.
MCP (Model Context Protocol): A protocol that exposes tools from external services to AI assistants through a standardized URL, allowing any MCP-compatible AI client to connect to a Composio MCP server without installing provider-specific packages.