Microsoft Teams API

What Microsoft Teams API does

Microsoft Teams API gives MCP-compatible agents access to Teams conversations and organizational people search. It covers one-to-one chats, group chats, meetings, channels, and their messages. Agents can locate conversations by topic, participant, or identifier, then read, send, edit, or delete messages.

The server also exposes profile lookup, conversation membership, the authenticated user's identity, message reactions, file downloads, and meeting transcript retrieval. Messages can include reactions, mentions, thread followers, and quoted-message references. Sending supports Markdown converted to rich HTML, image files, and document or media attachments.

How it works

The package runs as a Node.js MCP server through the teams-api-mcp command. A typical editor configuration launches it with npx, using the teams-api@latest package. MCP clients call named tools such as teams_list_conversations, teams_get_messages, and teams_send_message rather than using the CLI directly.

Most conversation operations accept one of three ways to identify a chat: a topic, a person's name for a one-to-one conversation, or a conversation ID. Channel replies additionally use the message ID of the post being replied to. One-to-one lookup can find an existing chat or create one when the person exists in the organization directory.

People and chat discovery use the Substrate search API when available. Authentication can be interactive, cached, based on a direct token, or handled through a debug session. macOS also supports FIDO2 auto-login; that mode is unavailable on Windows and Linux.

Setup and configuration

Node.js 18 or later is required. The server can be added through editor MCP configuration or launched with a command such as:

{
  "mcpServers": {
    "teams": {
      "command": "npx",
      "args": ["-y", "-p", "teams-api@latest", "teams-api-mcp"],
      "env": { "TEAMS_LOGIN": "true" }
    }
  }
}

TEAMS_LOGIN=true enables interactive authentication. TEAMS_AUTO can be used instead for unattended macOS authentication with a FIDO2 passkey. Credentials are cached in the macOS Keychain, Windows DPAPI or Credential Manager, and Linux secret storage. Image description requires an OpenAI-compatible vision service and either TEAMS_IMAGE_DESCRIPTION_API_KEY or OPENAI_API_KEY.

Tools and capabilities

• List conversations and search by topic, chat member, or person.
• Find or create one-to-one chats and retrieve Teams profiles.
• Read messages, channel thread replies, reactions, mentions, and quoted references.
• Send Markdown messages with inline images, file attachments, and image placeholders.
• Edit or delete messages and add or remove standard Teams reactions.
• List conversation members and identify the authenticated user.
• Retrieve parsed meeting transcripts or raw VTT files.
• Download message attachments, returning text and images inline when supported.
• Describe inline Teams images through a configured vision model.

Limitations and notes

One-to-one chats do not have topics, so they must be resolved by person name or direct conversation ID. If profile resolution is unavailable, one-to-one members may have empty display names. Binary downloads such as PDFs, videos, and archives are saved to disk rather than returned inline. Substrate-powered people and chat search depends on successful interactive or automatic authentication. Image description requires a separately configured compatible vision API key.