Ultimate Excel MCP Server

📸 Screenshots

(Replace these images with actual screenshots of your project before publishing)

Related MCP server: ironcalc-mcp

🌟 Introduction

Welcome to the Ultimate Excel MCP Server. This project bridges the gap between Advanced AI Agents (like Antigravity, Cursor, Claude) and your local Microsoft Excel application. By leveraging the Model Context Protocol (MCP), this server allows AI to read, write, format, and execute complex VBA macros directly inside your active Excel window in real-time.

Gone are the days of AI generating Python scripts for you to run manually, or exporting .csv files. With this server, you type a prompt, and the AI builds a fully functional spreadsheet right before your eyes.

🔥 Features (Super OP)

1. Live COM Integration: Connects directly to your open Microsoft Excel window using win32com. No need to save or close files.

2. Auto-Reconnection: If you close Excel and open it again, the server gracefully handles the COM disconnection and reconnects on the next polling cycle.

3. VBA Macro Injection (execute_vba_macro): The most OP feature. Allows AI to write and execute VBA macros on the fly to generate charts, pivot tables, custom UserForms, and advanced UIs in under a second.

4. Rich Formatting Control: Change cell background colors, font colors, bold/italic styles, and borders dynamically.

5. Offline Headless Mode: Includes an openpyxl engine for reading and writing .xlsx files silently in the background when Excel is not open.

6. Real-time React Dashboard: A beautiful, dark-mode web dashboard built with Vite + React + Tailwind v4 that monitors the AI's actions and logs all tool executions live via Server-Sent Events (SSE).

7. Zero JSON-RPC Corruption: Specifically engineered with standard output (stdout) redirection to stderr to prevent standard Python prints from breaking the MCP stdio protocol.

🏗️ Architecture & Flow

The system is composed of three main layers that communicate asynchronously to ensure lightning-fast performance:

flowchart TB
    classDef client fill:#2d3748,stroke:#4fd1c5,stroke-width:2px,color:#fff
    classDef server fill:#2b6cb0,stroke:#63b3ed,stroke-width:2px,color:#fff
    classDef target fill:#276749,stroke:#68d391,stroke-width:2px,color:#fff
    classDef dashboard fill:#702459,stroke:#d53f8c,stroke-width:2px,color:#fff

    A([🧠 AI Agent IDE]):::client <-->|JSON-RPC via stdio| B(🐍 FastAPI MCP Server):::server
    
    subgraph Engine Layer
    B -->|PyWin32 COM API| C[(📊 Live MS Excel)]:::target
    B -->|OpenPyxl| D[(📁 Offline .xlsx)]:::target
    end

    B -.->|SSE Logs & Metrics| E[💻 React Web Dashboard]:::dashboard
    E -.->|Axios REST API| B

📂 Directory Structure

excel-mcp-ai/
│
├── docs/                       # Project Images & Banners
├── mcp_server/                 # Python Backend
│   ├── main.py                 # FastAPI & MCP Entrypoint
│   ├── mcp_tools.py            # Definition of all 7 MCP Tools
│   ├── win32_engine.py         # Live COM controller for Excel
│   ├── offline_engine.py       # Openpyxl controller for files
│   ├── dashboard_api.py        # REST API & SSE for React
│   └── config.py               # Settings manager
│
├── dashboard/                  # React Frontend
│   ├── src/
│   │   ├── components/         # React Components
│   │   ├── App.tsx             # Main App
│   │   └── index.css           # Tailwind v4 directives
│   ├── package.json            
│   └── vite.config.ts          
│
├── requirements.txt            # Python dependencies
├── run.bat                     # Quick-start script for Windows
└── README.md                   # This file

🚀 Getting Started

Prerequisites

• Windows OS (Required for COM API)

• Microsoft Excel installed and activated

• Python 3.10+

• Node.js 18+ (For building the dashboard)

1. Installation

Clone the repository and set up the Python environment:

# Clone repo
git clone https://github.com/yourusername/ultimate-excel-mcp.git
cd ultimate-excel-mcp

# Create Virtual Environment
python -m venv venv
.\venv\Scripts\Activate.ps1

# Install Dependencies
pip install -r requirements.txt

2. Build the Dashboard

cd dashboard
npm install
npm run build
cd ..

3. Run the Server

To run the full suite (FastAPI Dashboard on Port 8000 + MCP stdio readiness):

.\run.bat

Note: Do NOT run run.bat in a restricted context. Allow PowerShell scripts if needed via Set-ExecutionPolicy RemoteSigned.

⚙️ MCP Configuration

To connect this server to your AI IDE (like Antigravity, Cursor, or Claude Desktop), you need to register it in your mcp_config.json file.

Crucial Step: Because the AI runs the command from arbitrary directories, you MUST set the PYTHONPATH environment variable so Python can resolve the mcp_server module.

Open your mcp_config.json and append:

{
  "mcpServers": {
    "Excel": {
      "command": "C:\\path\\to\\your\\excel-mcp-ai\\venv\\Scripts\\python.exe",
      "args": [
        "-m",
        "mcp_server.main",
        "--stdio"
      ],
      "env": {
        "PYTHONPATH": "C:\\path\\to\\your\\excel-mcp-ai"
      }
    }
  }
}

Replace C:\\path\\to\\your\\excel-mcp-ai with the absolute path to your cloned repository.

🛠️ API & Tools Reference

The server exposes 7 powerful MCP tools to the AI.

1. get_active_excel_status

Checks if Excel is running and retrieves the name of the currently active workbook and sheet.

• Parameters: None

• Returns: JSON object with connected, active_workbook, active_sheet.

• Note: If Excel is open but no workbook is active, this tool will automatically create a Blank Workbook (Workbooks.Add()).

2. write_active_excel_range

Injects 2D array data into the active sheet.

• Parameters:

  • range_address (str): e.g. "A1:C10"

  • data (list of lists): The 2D array of data.

  • sheet_name (str, optional): Target sheet. Defaults to active.

3. read_active_excel_range

Reads data from the active Excel sheet.

• Parameters:

  • range_address (str): e.g. "A1:C10"

  • sheet_name (str, optional): Target sheet.

4. format_active_excel_range

Applies aesthetic styling to a specific range.

• Parameters:

  • range_address (str)

  • options (dict):

    • bg_color (hex string, e.g., "#21A366")

    • font_color (hex string, e.g., "#FFFFFF")

    • bold (boolean)

5. execute_vba_macro (🌟 The OP Tool)

Injects and runs a Visual Basic for Applications (VBA) macro script directly in memory.

• Parameters:

  • macro_code (str): The full VBA code (e.g., Sub MyMacro() ... End Sub)

  • macro_name (str): The name of the subroutine to execute.

• Use Cases: Creating charts, generating pivot tables, turning off gridlines, conditional formatting, adding interactive buttons.

6. read_offline_excel

Reads data from a .xlsx file on disk using openpyxl without opening MS Excel.

• Parameters: file_path, range_address, sheet_name.

7. write_offline_excel

Writes data to a .xlsx file on disk silently.

• Parameters: file_path, range_address, data, sheet_name.

🤖 Prompts & AI Agent Skills

To maximize the potential of this server, you can give your AI specific "Skills". Here is an example of an OP Prompt you can feed to the AI:

"Build a completely automated Employee Attendance Dashboard. Use execute_vba_macro to create a new sheet called 'Dashboard', remove the gridlines, and insert a dark-blue header. Populate 15 lines of dummy attendance data with dynamic formulas computing the attendance percentage. Add a Bar Chart using VBA next to the table to visualize the data, and add a Macro-enabled 'Refresh' button."

The AI will parse this, write a complete VBA script on the fly, pass it to execute_vba_macro, and your Excel window will transform instantly.

🛡️ Trust Center Security Settings

For the execute_vba_macro tool to work, Microsoft Excel must allow programmatic access to the VBA object model.

1. Open Excel.

2. Go to File > Options > Trust Center > Trust Center Settings.

3. Select Macro Settings.

4. Check the box: ☑ Trust access to the VBA project object model.

5. Click OK.

🔧 Troubleshooting

1. invalid character 'I' looking for beginning of value

This error occurs in the IDE when the Python MCP server prints standard console text (like INFO: Server started) to stdout. The MCP protocol requires stdout to be strictly JSON. Fix: We have already routed all print() statements to sys.stderr in main.py and win32_engine.py. If you add custom prints, ensure you add , file=sys.stderr.

2. NameError: name 'CORSMiddleware' is not defined

Ensure that your main.py has the following import:

from fastapi.middleware.cors import CORSMiddleware

3. ModuleNotFoundError: No module named 'mcp_server'

Your AI IDE is executing the Python module from a different directory. You MUST set "env": { "PYTHONPATH": "..." } in your mcp_config.json as shown in the Configuration section.

4. COM Error: 'str' object is not callable

This happens in pywin32 when you try to call a property as a method. For example, ActiveCell.Address is a string in Python, not a callable function like it is in VBA. The backend handles this gracefully.

5. pywintypes.com_error: (-2147221008, 'CoInitialize has not been called.')

Because FastAPI routes run on separate async threads, win32com requires COM to be initialized on every thread. The win32_engine.py calls pythoncom.CoInitialize() at the start of every connection cycle to prevent this.

🤝 Contributing

Contributions make the open source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.

1. Fork the Project

2. Create your Feature Branch (git checkout -b feature/AmazingFeature)

3. Commit your Changes (git commit -m 'Add some AmazingFeature')

4. Push to the Branch (git push origin feature/AmazingFeature)

5. Open a Pull Request

📜 License

Distributed under the MIT License. See LICENSE for more information.