This MCP server allows the use of natural language to execute any CLI for Microsoft 365 commands. It may handle complex prompts that will be executed as a chain of CLI for Microsoft 365 commands that will try to fulfill the user's request. Thanks to this, you may manage many different areas of Microsoft 365, for example: Entra ID, OneDrive, OneNote, Outlook, Planner, Power Apps, Power Automate, Power Platform, SharePoint Embedded, SharePoint Online, Teams, Viva Engage, and many more...
For best results, use it with Claude Sonnet 4 or Claude Sonnet 3.7.
- Node.js 20.x or higher
- CLI for Microsoft 365 installed globally (
npm i -g @pnp/cli-microsoft365
)
๐ Check out the Getting Started recording ๐ฌ
This MCP server uses the globally installed CLI for Microsoft 365 that you need to install globally using npm i -g @pnp/cli-microsoft365
.
After you install the CLI for Microsoft 365 perform the initial setup by running the:
m365 setup
For more information, please follow the Log in to Microsoft 365
After you set up the CLI for Microsoft 365, please update its configuration using the following commands:
m365 cli config set --key prompt --value false
m365 cli config set --key output --value text
m365 cli config set --key helpMode --value full
This will ensure that the MCP server will get as much information as possible from the CLI for Microsoft 365 when an error occurs or when a command is not executed properly.
The MCP server will not do any authentication for you. You will need to first authenticate using CLI for Microsoft 365 using m365 login command. Once you are authenticated, the MCP server will use the same authentication context when running any tool.
Lastly, install the MCP server on your preferred MCP client, referring to the instructions of the picked MCP host. Below, you may find guidance for VS Code.
There are multiple ways to add the MCP server to VS Code. The most common way is to:
-
Open the Command Palette (Ctrl+Shift+P or Cmd+Shift+P on macOS) and type
MCP: Add Server
. -
Select
Command (stdio)
as the server type. -
Enter the command to run the MCP server, which is:
npx -y @pnp/cli-microsoft365-mcp-server@latest
-
Name the server (e.g.,
CLI for Microsoft 365 MCP Server
).
As a result, you should have the following configuration in your .vscode/mcp.json
file:
{
"servers": {
"CLI for Microsoft 365 MCP Server": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@pnp/cli-microsoft365-mcp-server@latest"
]
}
}
}
For best results, use it with Claude Sonnet 4 or Claude Sonnet 3.7.
The below uses cases are only a few example how you may use this MCP server. It is capable of handling many different tasks, so feel free to experiment with it and basically manage all of Microsoft 365 using natural language.
Also you may check the CLI for Microsoft 365 wiki to find more scenarios how ClI for Microsoft 365 can be used to manage Microsoft 365.
In the following example, the MCP server created a SharePoint list with new columns and populated it with a few test items.
prompt: "Add a new list to this site with title "awesome ducks". Then add new columns to that list including them in the default view. The first should be a text description columns and the second one should be a user column. Then add 3 items to this list with some funny jokes about docs added in the description column and adding my user in the user column. use emojis ๐"
The following example shows how to create a new Microsoft Teams team and add a welcome post to the General channel.
prompt: "Create a new Team on Teams with name "Awesome Ducks" and in the General channel adda a welcome post"
In the example, the MCP server disables a Power Automate flow.
prompt: "can you check if I have HoursReportingReminder flow and if so disable it"
In the following example, the MCP server creates a new Planner plan with sample buckets and tasks.
prompt: "can you create a new plan in planner to manage work for the awesome ducks. I need some sample buckets and tasks to get started"
Column 1 | Column 2 |
---|---|
m365GetCommands | Gets all CLI for Microsoft 365 commands to be used by the Model Context Protocol to pick the right command for a given task |
m365GetCommandDocs | Gets documentation for a specified CLI for Microsoft 365 command to be used by the Model Context Protocol to provide detailed information about the command along with examples, use cases, and option descriptions |
m365RunCommand | Runs a specified CLI for Microsoft 365 command to be used by the Model Context Protocol to execute the command and return the result and reason over the response |
Before anything, first run npm install
to install all dependencies.
Then, in order to build the project run:
npm run build
It is also possible to run the MCP server in VS Code from your local build so that it may be used by GitHub Copilot Agent. First, start the CLI for Microsoft 365 MCP server using the command:
npm run start
Now go to VS Code GitHub Copilot Agent mode click on the tools icon, and select Add more tools
. Then select Add MCP server
and then Command (stdio)
and enter the following command:
node FULL_PATH_TO_YOUR_PROJECT/dist/index.js
Click Enter and name it however you like. It is recommended to add it to workspace
scope for testing. After that, open up your .vscode/mcp.json
file and modify it so pass the environment variables needed for auth.
{
"servers": {
"m365-mcp-server": {
"type": "stdio",
"command": "node",
"args": [
"FULL_PATH_TO_YOUR_PROJECT/dist/index.js" // e.g. C:/workspace/repo/microsoft-365-mcp-server/dist/index.js
]
}
}
}
One of the ways to test the CLI for Microsoft 365 MCP server is by using the MCP Inspector. Mostly, this setup is used for debugging purposes and to test the MCP server locally. First, start the MCP server using the command:
npm run start
Now, in order to run the inspector for your MCP server, you need run the following command in the repo root folder location:
npx @modelcontextprotocol/inspector node dist/index.js
After that, wait for the inspector to start and open the inspector in your browser. You should see the MCP server running, and you should be able to query the tools and execute them locally.