home / mcp / aisecretary mcp server

AISecretary MCP Server

Modular MCP servers for automating Outlook tools (Mail, Calendar, Contacts, To Do, Mailbox Settings, Categories)

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "diegodopazo-aisecretary": {
      "command": "/Your/path/to/uv",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "--with",
        "msal",
        "mcp",
        "run",
        "/your/path/to/server.py"
      ]
    }
  }
}

AISecretary MCP Server provides modular, MCP-based automation for Outlook-related tools. Each server handles a specific Outlook task set (Mail, Calendar, Contacts, To Do, Mailbox Settings, Categories), enabling you to start only what you need and integrate with platforms like OpenWebUI or Claude Desktop.

How to use

You interact with AISecretary through an MCP client or an integration platform that supports MCP servers. Start the specific MCP server you want (for example, Email or Calendar) and connect the client to the local MCP endpoint. Once connected, issue tasks like checking emails, scheduling events, managing contacts, or organizing folders using natural prompts or predefined tool actions. You can run multiple servers in parallel to cover different Outlook tools and combine their responses to automate daily workflows.

To enable OpenWebUI or Claude Desktop integrations, ensure you have the corresponding runtime commands running, then point the UI to the local MCP endpoint to begin interacting with AISecretary’s capabilities. Use the UI’s Tools section to add the local address where MCP is exposed, typically on localhost with a specific port, and start issuing commands to manage Outlook data. Ensure your authentication flow (e.g., Microsoft Graph) is configured for the required scopes.

How to install

Prerequisites: install uv in your Python environment and ensure Python 3.11 or higher is available.

Step 1: Install dependencies if needed and prepare your Python environment. Ensure you can run Python scripts with the uv tool.

Step 2: Start the MQTT-like MCP server(s) you want to use. Use the provided command patterns to launch the server files (server.py or equivalent) with uv or uvx as shown in the sample configurations.

Step 3: If you are using Claude Desktop, start the server with the MCP run command and ensure it registers under your Claude setup. For example, you can run the install command to register the server with Claude and then start the server file.

Additional configuration and start examples

Two explicit configuration examples are shown to guide setup in different environments. These examples configure an MCP server entry that runs the runtime via uv and points to your server script.

{
  "mcpServers": {
    "server_name": {
      "command": "/Your/path/to/uv",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "--with",
        "msal",
        "mcp",
        "run",
        "/your/path/to/server.py"
      ]
    }
  }
}

If you are using an OpenWebUI setup, you can also start the MCP server with a direct command and then register the local endpoint in the UI.

{
  "mcpServers": {
    "openwebui_server": {
      "command": "uvx",
      "args": [
        "mcpo",
        "--port",
        "9000",
        "--",
        "uv",
        "run",
        "mcp",
        "run",
        "server.py"
      ]
    }
  }
}

Security and maintenance notes

Configure Microsoft Graph API access securely and provide only the necessary scopes for Outlook operations. Use environment isolation for credentials and rotate tokens regularly. Keep the MCP servers updated with the latest compatible dependencies to ensure security and stability.

Troubleshooting tips

If a server fails to start, verify the path to your uv/uvx runtime, confirm the server file path, and check that the port you are binding to is available. Ensure your environment variables for Graph authentication are set and that your OpenWebUI or Claude Desktop integration is configured to reach the correct local endpoint.

Notes

AISecretary is designed to be modular. Start only the Outlook tools you need and add new servers as you expand automation coverage. All changes to server configurations should be tested in a controlled environment before deploying to production.

Available tools

Email

Advanced email search, retrieve conversations, manage folders, create/edit drafts, send replies, attachments, rules, and more.

Categories

View, create, edit, delete categories and assign colors; apply categories to emails and events.

Calendar

Manage events: create, update, delete; handle invitations; manage attachments and calendars.

Task Management

Manage task lists and tasks: create, retrieve, update, delete.

Contacts

Manage contact folders and contacts: create, search, retrieve details, update, delete.

Mailbox Settings

Retrieve and update general mailbox settings including automatic replies, time zone, and working hours.