Teams MCP Server
Enables AI assistants to interact with Microsoft Teams through direct API calls, supporting message search, sending replies, managing favorites, and more without Azure AD registration.
README
Teams MCP Server
An MCP (Model Context Protocol) server that enables AI assistants to interact with Microsoft Teams. Search messages, send replies, manage favourites, and more.
How It Works
This server calls Microsoft's Teams APIs directly (Substrate, chatsvc, CSA) - the same APIs the Teams web app uses. No Azure AD app registration or admin consent required.
Authentication flow:
- AI runs
teams_loginto open a browser for you to log in - OAuth tokens are extracted and cached
- All operations use cached tokens directly (no browser needed)
- Automatic token refresh (~1 hour)
Security: Uses the same authentication as the Teams web client - your access is limited to what your account can already do.
Installation
Prerequisites
- Node.js 18+
- A Microsoft account with Teams access
- Google Chrome, Microsoft Edge, or Chromium browser installed
Configure Your MCP Client
Add to your MCP client configuration (e.g., Claude Desktop, Windsurf, Cursor):
{
"mcpServers": {
"teams": {
"command": "npx",
"args": ["-y", "msteams-mcp@latest"]
}
}
}
That's it. npx will automatically download and run the latest version.
From Source (alternative)
If you prefer to run from a local clone:
git clone https://github.com/m0nkmaster/msteams-mcp.git
cd msteams-mcp
npm install && npm run build
Then configure your MCP client:
{
"mcpServers": {
"teams": {
"command": "node",
"args": ["/path/to/msteams-mcp/dist/index.js"]
}
}
}
The server uses your system's Chrome (macOS/Linux) or Edge (Windows) for authentication.
Available Tools
Search & Discovery
| Tool | Description |
|---|---|
teams_search |
Search Teams messages with operators (from:, sent:, in:, hasattachment:, etc.) |
teams_search_email |
Search emails in your mailbox (same auth as Teams — no extra login) |
teams_get_message |
Get a single message by ID with full content (any age) |
teams_get_thread |
Get messages from a conversation/thread |
teams_find_channel |
Find channels by name (your teams + org-wide discovery) |
teams_get_activity |
Get activity feed (mentions, reactions, replies, notifications) |
Messaging
| Tool | Description |
|---|---|
teams_send_message |
Send a message (default: self-chat/notes). Use replyToMessageId for thread replies |
teams_edit_message |
Edit one of your own messages |
teams_delete_message |
Delete one of your own messages (soft delete) |
People & Contacts
| Tool | Description |
|---|---|
teams_get_me |
Get current user profile (email, name, ID) |
teams_search_people |
Search for people by name or email |
teams_get_frequent_contacts |
Get frequently contacted people (useful for name resolution) |
teams_get_chat |
Get conversation ID for 1:1 chat with a person |
teams_create_group_chat |
Create a new group chat with multiple people (2+ others) |
Organisation
| Tool | Description |
|---|---|
teams_get_favorites |
Get pinned/favourite conversations |
teams_add_favorite |
Pin a conversation |
teams_remove_favorite |
Unpin a conversation |
teams_save_message |
Bookmark a message |
teams_unsave_message |
Remove bookmark from a message |
teams_get_saved_messages |
Get list of saved/bookmarked messages with source references |
teams_get_followed_threads |
Get list of followed threads with source references |
teams_get_unread |
Get unread counts (aggregate or per-conversation) |
teams_mark_read |
Mark a conversation as read up to a message |
Reactions
| Tool | Description |
|---|---|
teams_search_emoji |
Search for emojis by name (standard + custom org emojis) |
teams_add_reaction |
Add an emoji reaction to a message |
teams_remove_reaction |
Remove an emoji reaction from a message |
Quick reactions: like, heart, laugh, surprised, sad, angry can be used directly without searching.
Calendar & Meetings
| Tool | Description |
|---|---|
teams_get_meetings |
Get meetings from calendar (defaults to next 7 days) |
teams_get_transcript |
Get meeting transcript (requires threadId from teams_get_meetings) |
teams_get_meetings returns: subject, times, organiser, join URL, threadId for meeting chat. Use threadId with teams_get_thread to read meeting chat, or with teams_get_transcript to get the full transcript with speakers and timestamps.
Files
| Tool | Description |
|---|---|
teams_get_shared_files |
Get files and links shared in a conversation (supports pagination) |
Returns both files (name, extension, URL, size) and links (URL, title), along with who shared each item. Works for channels, group chats, 1:1 chats, and meeting chats.
Session
| Tool | Description |
|---|---|
teams_login |
Trigger manual login (opens browser) |
teams_status |
Check authentication and session state |
Search Operators
Both teams_search (Teams messages) and teams_search_email (emails) support native operators:
from:sarah@company.com # Messages/emails from person
sent:2026-01-20 # From specific date
sent:>=2026-01-15 # Since date
in:project-alpha # Messages in channel (Teams only)
subject:"budget" # By subject (email)
"Rob Smith" # Find @mentions (name in quotes)
hasattachment:true # With files
is:unread # Unread emails (email only)
NOT from:email@co.com # Exclude results
Combine operators: from:sarah@co.com sent:>=2026-01-18 hasattachment:true
Note: @me, from:me, to:me do NOT work. Use teams_get_me first to get your email/displayName. sent:today works, but sent:lastweek and sent:thisweek do NOT - use explicit dates or omit (results are sorted by recency).
MCP Resources
The server also exposes passive resources for context discovery:
| Resource URI | Description |
|---|---|
teams://me/profile |
Current user's profile |
teams://me/favorites |
Pinned conversations |
teams://status |
Authentication status |
CLI Tools (Development)
For local development, CLI tools are available for testing and debugging:
# Check authentication status
npm run cli -- status
# Search messages
npm run cli -- search "meeting notes"
npm run cli -- search "project" --from 0 --size 50
# Search emails
npm run cli -- teams_search_email --query "from:sarah@company.com"
# Send messages (default: your own notes/self-chat)
npm run cli -- send "Hello from Teams MCP!"
npm run cli -- send "Message" --to "conversation-id"
# Force login
npm run cli -- login --force
# Output as JSON
npm run cli -- search "query" --json
MCP Test Harness
Test the server through the actual MCP protocol:
# List available tools
npm run cli
# Call any tool
npm run cli -- search "your query"
npm run cli -- status
npm run cli -- people "john smith"
npm run cli -- favorites
npm run cli -- activity # Get activity feed
npm run cli -- unread # Check unread counts
npm run cli -- teams_search_emoji --query "heart" # Search emojis
Limitations
- Login required - Run
teams_loginto authenticate (opens browser) - Token expiry - Tokens expire after ~1 hour; headless refresh is attempted or run
teams_loginagain when needed - Undocumented APIs - Uses Microsoft's internal APIs which may change without notice
- Search limitations - Full-text search only; thread replies not matching search terms won't appear (use
teams_get_threadfor full context) - Own messages only - Edit/delete only works on your own messages
Session Files
Session files are stored in a user config directory (encrypted):
- macOS/Linux:
~/.teams-mcp-server/ - Windows:
%APPDATA%\teams-mcp-server\
Contents: session-state.json, token-cache.json, browser-profile/
If your session expires, call teams_login or delete the config directory.
Development
For local development:
git clone https://github.com/m0nkmaster/msteams-mcp.git
cd msteams-mcp
npm install
npm run build
Development commands:
npm run dev # Run MCP server in dev mode
npm run build # Compile TypeScript
npm run lint # Run ESLint
npm run research # Explore Teams APIs (logs network calls)
npm test # Run unit tests
npm run typecheck # TypeScript type checking
For development with hot reload, configure your MCP client:
{
"mcpServers": {
"teams": {
"command": "npx",
"args": ["tsx", "/path/to/msteams-mcp/src/index.ts"]
}
}
}
See AGENTS.md for detailed architecture and contribution guidelines.
Teams Chat Export Bookmarklet
This repo also includes a standalone bookmarklet for exporting Teams chat messages to Markdown. See teams-bookmarklet/README.md.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。