microsoft-agent 0.2.56

Microsoft Graph Agent MCP Server

Microsoft Agent - A2A | AG-UI | MCP

Version: 0.2.56

Overview

Microsoft Graph MCP Server + A2A Supervisor Agent

It includes a Model Context Protocol (MCP) server that wraps the Microsoft Graph API and an out-of-the-box Agent2Agent (A2A) Supervisor Agent.

Manage your Microsoft 365 tenant (Users, Groups, Calendars, Drive, etc.) through natural language!

This repository is actively maintained - Contributions are welcome!

Capabilities:

• Comprehensive Graph API Coverage: Access thousands of Microsoft Graph endpoints via MCP tools.
• Supervisor-Worker Agent Architecture: A smart supervisor delegates tasks to specialized agents (e.g., Calendar Agent, User Agent).
• Secure Authentication: Supports OAuth, OIDC, and other authentication methods.

MCP

MCP Tools

This server provides tools for a vast array of Microsoft Graph resources. Due to the large number of tools, they are organized by resource type.

Supported Resources (Partial List):

• Users: get_user, update_user, list_user, etc.
• Groups: get_group, post_groups_group, list_members_group, etc.
• Calendar: get_calendar, post_events, list_calendarview, etc.
• Drive/DriveItems: get_drive, search_driveitem, upload_driveitem, etc.
• Mail: send_mail, list_messages, etc.
• Directory Objects: get_directoryobject, check_member_objects, etc.
• Planner, OneNote, Teams, and more.

All tools generally follow the naming convention: action_resource (e.g., list_user, delete_group).

Using as an MCP Server

The MCP Server can be run in two modes: stdio (for local testing) or http (for networked access).

Run in stdio mode (default):

microsoft-agent --transport "stdio"

Run in HTTP mode:

microsoft-agent --transport "http"  --host "0.0.0.0"  --port "8000"

AI Prompt:

Find who manages the 'Engineering' group and list its members.

AI Response:

I've found the 'Engineering' group.
Owners:
- Jane Doe (Head of Engineering)

Members:
- John Smith
- Alice Johnson
- Bob Williams
...

A2A Agent

This package includes a powerful A2A Supervisor Agent that orchestrates interaction with the Microsoft MCP tools.

Architecture

The system uses a Supervisor Agent that analyzes user requests and delegates them to domain-specific Child Agents.

---
config:
  layout: dagre
---
flowchart TB
 subgraph subGraph0["Agent System"]
        S["Supervisor Agent"]
        subGraph1["Specialized Agents"]
            CA["Calendar Agent"]
            GA["Group Agent"]
            UA["User Agent"]
            DA["Drive Agent"]
            OA["...Other Agents"]
        end
        B["A2A Server"]
        M["MCP Tools"]
  end
    U["User Query"] --> B
    B --> S
    S --Delegates--> CA & GA & UA & DA & OA
    CA & GA & UA & DA & OA --> M
    M --> api["Microsoft Graph API"]

     S:::agent
     B:::server
     U:::server
     CA:::worker
     GA:::worker
     UA:::worker
     DA:::worker
     OA:::worker

    classDef server fill:#f9f,stroke:#333
    classDef agent fill:#bbf,stroke:#333,stroke-width:2px
    classDef worker fill:#dcedc8,stroke:#333
    style B stroke:#000000,fill:#FFD600
    style M stroke:#000000,fill:#BBDEFB
    style U fill:#C8E6C9
    style subGraph0 fill:#FFF9C4

Component Interaction

1. User sends a request (e.g., "Schedule a meeting with the Engineering team").
2. Supervisor Agent identifies this as a calendar and group task.
3. Supervisor delegates finding the group members to the Group Agent.
4. Group Agent calls list_members_group tool and returns emails.
5. Supervisor delegates scheduling to the Calendar Agent with the retrieved emails.
6. Calendar Agent calls post_events tool.
7. Supervisor confirms completion to the User.

Graph Architecture

This agent uses pydantic-graph orchestration for intelligent routing and optimal context management.

---
title: Microsoft Agent Graph Agent
---
stateDiagram-v2
  [*] --> RouterNode: User Query
  RouterNode --> DomainNode: Classified Domain
  RouterNode --> [*]: Low confidence / Error
  DomainNode --> [*]: Domain Result

• RouterNode: A fast, lightweight LLM (e.g., nvidia/nemotron-3-super) that classifies the user's query into one of the specialized domains.
• DomainNode: The executor node. For the selected domain, it dynamically sets environment variables to temporarily enable ONLY the tools relevant to that domain, creating a highly focused sub-agent (e.g., gpt-4o) to complete the request. This preserves LLM context and prevents tool hallucination.

Usage

MCP CLI

Short Flag	Long Flag	Description
-h	--help	Display help information
-t	--transport	Transport method: 'stdio', 'http', or 'sse' [legacy] (default: stdio)
-s	--host	Host address for HTTP transport (default: 0.0.0.0)
-p	--port	Port number for HTTP transport (default: 8000)
	--auth-type	Auth type: 'none', 'static', 'jwt', 'oauth-proxy', 'oidc-proxy' (default: none)
	...	(See standard FastMCP auth flags)

A2A CLI

Endpoints

• Web UI: http://localhost:9000/ (if enabled)
• A2A: http://localhost:9000/a2a (Discovery: /a2a/.well-known/agent.json)
• AG-UI: http://localhost:9000/ag-ui (POST)

Argument	Description	Default
--host	Host to bind the server to	0.0.0.0
--port	Port to bind the server to	9000
--provider	LLM Provider (openai, anthropic, google, huggingface)	openai
--model-id	LLM Model ID	nvidia/nemotron-3-super
--mcp-url	MCP Server URL	http://microsoft-agent:8000/mcp

Examples

Run A2A Server

microsoft-agent-server --provider openai --model-id gpt-4o --api-key sk-... --mcp-url http://localhost:8000/mcp

Docker

Build

docker build -t microsoft-agent .

Run MCP Server

docker run -p 8000:8000 microsoft-agent

Run Agent Server

docker run -e CMD=agent-server -p 9000:9000 microsoft-agent

Deploy as a Service

docker pull knucklessg1/microsoft-agent:latest

docker run -d \
  --name microsoft-agent \
  -p 8000:8000 \
  -e HOST=0.0.0.0 \
  -e PORT=8000 \
  -e TRANSPORT=http \
  knucklessg1/microsoft-agent:latest

Install Python Package

python -m pip install microsoft-agent

uv pip install microsoft-agent

Repository Owners

Documentation:

Microsoft API Docs Microsoft Graph SDK