Is Outlook MCP Server an official Microsoft product?
No. It is a community project by GitHub user Wallisking1991. Microsoft does not publish or support it.
@Wallisking1991
Basic information
No overview available yet
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"outlook-mcp-server": {
"command": "python",
"args": [
"outlook_mcp_server.py"
]
}
}
}Tool list not available yet
This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.
We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.
Outlook MCP Server is a community outlook mcp server for Windows that lets Claude and other MCP clients list, search, read, reply to and send email through your locally installed Microsoft Outlook.
It is a single Python script (outlook_mcp_server.py) built on the MCP Python SDK's FastMCP and pywin32. Instead of calling a cloud API, it drives the Outlook desktop application over COM/MAPI, so it works with whatever account Outlook is already signed into. It is maintained on GitHub by Wallisking1991 and is not an official Microsoft product.
pywin32 and the Outlook COM interface)No API key, OAuth app or Microsoft Graph registration is required. The server uses the Outlook profile on your machine.
Clone or download the repository from GitHub, then install the dependencies:
pip install mcp>=1.2.0 pywin32>=305
The same two packages are listed in the repo's requirements.txt. You can start the server by hand to check that it can reach Outlook:
python outlook_mcp_server.py
On startup it connects to Outlook and prints how many items are in your Inbox before starting the MCP server. In normal use you let your MCP client launch it instead.
Add the server to your Claude Desktop MCP configuration, replacing the path with the full path to the script on your machine:
{
"mcpServers": {
"outlook": {
"command": "python",
"args": ["Your path\\outlook_mcp_server.py"],
"env": {}
}
}
}
Restart Claude Desktop after saving. The repo ships a sample claude_desktop_config.json that uses an absolute Windows path such as d:\\Cline\\MCP\\outlook_mcp_server.py. Note the doubled backslashes required in JSON.
The server runs over standard input/output, so any client that can launch a local command (Cursor, VS Code, Claude Code and others) can use the same command and args. The README only documents the Claude Desktop setup, so use your client's own instructions for registering a local stdio server.
OR between terms.The README documents six tools. They fall into three groups.
Browsing and search. list_folders returns the folders in your mailbox. list_recent_emails lists email titles from a number of days (default 7, maximum 30) and optionally takes a folder name. search_emails takes a search term plus the same days and folder options, and supports OR queries such as project update OR meeting notes.
Reading. get_email_by_number returns the full content of an email using the number it was given in the most recent list_recent_emails or search_emails result. You need to list or search first, otherwise the tool returns an error.
Writing. reply_to_email_by_number replies to a numbered email with your text. compose_email takes a recipient, subject, body and optional CC address, then sends the message straight away rather than saving a draft.
A typical session runs list_folders, then list_recent_emails or search_emails, then get_email_by_number, and finally reply_to_email_by_number.
This outlook mcp server is local and Windows-only. It controls desktop Outlook through pywin32, so it needs no Microsoft 365 app registration, but it cannot run on macOS, Linux or a remote host. It also covers email only.
Microsoft Graph-based Outlook MCP servers take the other approach: they connect to Microsoft 365 mailboxes in the cloud and often cover calendar and contacts too. The trade-off is OAuth or app setup. One hosted example is Pipedream's Microsoft Outlook MCP. If your mail is in Gmail rather than Outlook, see Gmail MCP Server.
python outlook_mcp_server.py by hand shows whether the script can reach your Inbox.get_email_by_number and reply_to_email_by_number work from the last listing. Run list_recent_emails or search_emails first.OR between terms. Results depend on Outlook's built-in search.No. It is a community project by GitHub user Wallisking1991. Microsoft does not publish or support it.
Yes. It is a free Python script published on GitHub, with no paid tier. You do need a working Microsoft Outlook installation and account.
No. The server talks to the Outlook desktop app on your PC through `pywin32`, so no API key, OAuth app or Graph permission is needed.
No. It requires Windows because it relies on `pywin32` and the Outlook COM interface. On other systems, use a Microsoft Graph-based server instead.
The README documents Claude Desktop, but the server is a standard local stdio MCP server. Any client that can run `python` with the script path as an argument should work.
No. It only supports basic email functions: folders, listing, search, reading, replying and composing. Calendar and contacts are out of scope.
Up to 30 days. The `days` parameter on `list_recent_emails` and `search_emails` must be between 1 and 30.
The server can read, send and reply to email from your Outlook account, and `compose_email` sends immediately. The README advises using it only with trusted MCP clients in secure environments. Review tool calls before you approve them.

Give your AI agent a phone. JooVoice makes real calls for you in 80+ languages, with local accents, across 67 countries, like booking a table, moving an appointment or chasing a refund, and sends back what happened. Call

Run a brand's social media from your agent: write posts in the brand's voice, generate short videos, schedule and publish to TikTok, Instagram, YouTube, X, LinkedIn, Facebook and Pinterest, and read the metrics back. 21
SendRaven is email infrastructure for AI agents. Send, read replies as threads, run campaigns and automations, with limits on the API key.

AI call analysis and voice agents for sales teams. This MCP server lets a team's own AI assistant read minute usage, agents, call summaries and daily stats, look up caller history, schedule consent-gated AI callbacks, an

Agencies and GTM teams use HeyReach and Instantly to run LinkedIn outbound from personal sender profiles — but those profiles sit empty and prospects don't accept. B2B Creators is the content layer that keeps every sende
Comments