m365-helping-agent

m365-helping-agent

Enables managing car repair records through MCP tools, with OAuth authentication via Microsoft Entra ID.

Category
访问服务器

README

HelpingAgent

A learning project demonstrating how to connect a Microsoft 365 Declarative Agent to a remote Model Context Protocol (MCP) server secured with Microsoft Entra ID OAuth authentication.


Overview

HelpingAgent is a Microsoft 365 Copilot Declarative Agent that helps users manage car repair records. Instead of calling a traditional REST API directly, the agent communicates with a remote MCP server over Streamable HTTP. Every tool invocation is protected using OAuth 2.0 with Microsoft Entra ID, ensuring that only authenticated users can access repair data.

The project uses Azure Table Storage (emulated locally using Azurite) as the backend datastore and demonstrates an end-to-end implementation of:

  • Microsoft 365 Declarative Agents
  • Model Context Protocol (MCP)
  • OAuth authentication with Microsoft Entra ID
  • Azure Table Storage
  • Microsoft 365 Agents Toolkit

Features

Declarative Agent

  • List repair records
  • View repair details
  • Create new repairs
  • Update existing repairs
  • Filter repairs by assigned technician

Remote MCP Server

  • Built using FastMCP
  • Uses Streamable HTTP transport
  • Exposes repair operations as MCP tools
  • Communicates securely with Microsoft 365 Copilot

OAuth Authentication

  • Microsoft Entra ID issued Bearer (JWT) tokens
  • JWKS-based signature verification
  • Validation of:
    • Issuer
    • Audience
    • Required scopes

Data Storage

  • Azure Table Storage
  • Local development using Azurite

Architecture

                Microsoft 365 Copilot
              (Declarative Agent)

                       │
                       │
      OAuth Protected MCP Tool Calls
         (Bearer JWT Access Token)
                       │
                       ▼

          MCP Server (FastMCP)
        Streamable HTTP Transport
        src/mcp_server.py

      • JWT Signature Validation
      • Issuer Validation
      • Audience Validation
      • Scope Validation

                       │
                       ▼

          Azure Table Storage
           Repairs Table
      (Azurite for Local Development)

Repository Structure

Path Description
src/mcp_server.py MCP server implementation containing repair tools and OAuth token validation.
src/init_data.py Seeds Azure Table Storage with sample repair records.
src/repairsData.json Sample repair data used during initialization.
appPackage/repairDeclarativeAgent.json Declarative Agent definition.
appPackage/ai-plugin.json Plugin configuration connecting the agent to the remote MCP server.
appPackage/instruction.txt System instructions for the Declarative Agent.
appPackage/adaptiveCards/ Adaptive Card templates used to display repair information.
appPackage/manifest.json Microsoft Teams/Microsoft 365 app manifest.
infra/azure.bicep Azure infrastructure definition.
m365agents.yml Microsoft 365 Agents Toolkit provisioning and deployment pipeline.
m365agents.local.yml Local development pipeline.
env/ Environment configuration files.
mcp-auth-todo.txt OAuth implementation checklist.
docs/HelpingAgent-Learning-Guide.pdf Project learning guide.

OAuth Authentication Flow

The MCP server is secured using Microsoft Entra ID OAuth.

  1. A Microsoft Entra ID application exposes the custom scope:

    access_as_user
    
  2. The MCP server is configured with:

    • OAUTH_ACCEPTED_ISSUERS
    • OAUTH_ACCEPTED_AUDIENCES
    • OAUTH_REQUIRED_SCOPES
    • OAUTH_JWKS_URI
  3. For every MCP tool call:

    • Extract the Bearer token
    • Verify the JWT signature using Microsoft's JWKS
    • Validate the issuer
    • Validate the audience
    • Validate the required scope
    • Reject unauthorized requests
  4. The plugin (ai-plugin.json) uses:

    "auth": {
      "type": "OAuthPluginVault"
    }
    

    Microsoft 365 Copilot handles user authentication, consent, token acquisition, and automatically includes the Bearer token with every MCP request.

  5. During provisioning, the Microsoft 365 Agents Toolkit registers the OAuth configuration using the oauth/register step and stores the generated reference ID in the environment configuration for use by the plugin.


Prerequisites

  • Python 3.13
  • Node.js
  • Microsoft 365 Agents Toolkit
  • Microsoft 365 Copilot enabled tenant
  • Microsoft Entra ID App Registration
  • Azure Table Storage (Azurite for local development)

Local Development

Install dependencies

pip install -r requirements.txt
npm install

Start Azurite

npm run start:azurite

Seed sample data

npm run init-data

Start the MCP Server

npm run start:mcp-http

The MCP server runs at:

http://127.0.0.1:3001

Inspect available MCP tools

npm run inspector

Run the legacy HTTP Azure Function (optional)

npm run dev

Run the complete solution

Press F5 in Visual Studio Code using the Microsoft 365 Agents Toolkit.

The toolkit will:

  • Provision the application
  • Start the MCP server
  • Create a Development Tunnel
  • Configure OAuth
  • Sideload the Declarative Agent into Microsoft 365 Copilot

Environment Configuration

Environment variables are organized by deployment stage.

env/
├── .env.local
├── .env.local.user
├── .env.dev
└── .env.dev.user
  • .env.* files contain non-secret configuration.
  • .env.*.user files contain secrets and are excluded from source control.

Important Variables

Variable Description
AZURE_STORAGE_CONNECTION_STRING Azure Table Storage connection string.
OAUTH_ACCEPTED_ISSUERS Accepted Microsoft Entra ID issuer(s).
OAUTH_ACCEPTED_AUDIENCES Accepted App ID URI(s).
OAUTH_REQUIRED_SCOPES Required OAuth scope(s).
OAUTH_JWKS_URI Microsoft Entra ID JWKS endpoint.
MCP_SERVER_URL Public MCP server URL.

Technologies Used

  • Microsoft 365 Copilot
  • Microsoft 365 Declarative Agents
  • Model Context Protocol (MCP)
  • FastMCP
  • Streamable HTTP
  • Microsoft Entra ID
  • OAuth 2.0
  • JSON Web Tokens (JWT)
  • JSON Web Key Sets (JWKS)
  • Azure Table Storage
  • Azurite
  • Microsoft 365 Agents Toolkit
  • Python 3.13

Project Status

  • ✅ Microsoft 365 Declarative Agent implemented
  • ✅ Remote MCP Server implemented
  • ✅ Azure Table Storage integration completed
  • ✅ OAuth authentication with Microsoft Entra ID implemented
  • ✅ JWT validation using JWKS implemented
  • ✅ Streamable HTTP transport configured
  • ✅ End-to-end communication between Microsoft 365 Copilot and the MCP server verified

Learning Objectives

This project was built as a hands-on learning exercise to understand:

  • Building Microsoft 365 Declarative Agents
  • Developing MCP servers using FastMCP
  • Exposing functionality as MCP tools
  • Securing MCP servers with Microsoft Entra ID OAuth
  • Validating JWT access tokens using JWKS
  • Connecting Microsoft 365 Copilot to remote MCP servers
  • Using Azure Table Storage as the backend datastore
  • Provisioning and deploying applications with the Microsoft 365 Agents Toolkit

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选