Google Chat MCP Server

Google Chat MCP

English | Русский

A1 Google Chat MCP lets an AI app work in Google Chat in plain language. Find the right space or direct message, catch up on a conversation, reply in a thread, react with an emoji and manage who is in a space.

It uses the Google Chat API with your Google account and acts as the signed-in user: messages send under your name, and only your own messages and reactions can be edited or deleted. It makes the limits of the Chat API explicit instead of implying that every chat task is possible.

• 20 tools. Connect from the chat, discover spaces and direct messages, read and send messages with thread control, manage emoji reactions, read attachment metadata and manage membership.
• Connects from the conversation. Say "connect Google Chat": the server walks you through the OAuth client, catches Google's redirect on 127.0.0.1 with PKCE and keeps the tokens itself — no config files, no restart.
• You act as yourself. Sends appear under your name; edits and deletes stop at your own messages and reactions.
• A send is never replayed. After an ambiguous failure the server does not retry a write — a replayed send would be a duplicate message in a real room.
• Minimal Google scopes. The server sends whatever token you minted; grant scopes per task — read-only ones are enough for browsing spaces and messages.

Start with a read-only question:

Show today’s messages in the team space and summarize what was decided.

Connect the server · Explore use cases · Open technical documentation

---

See it work in a minute

You: What was discussed in the release space today?

Assistant: Lists today’s messages with senders and threads. Nothing changes.

You: Reply in the deploy thread that the rollout is finished.

Assistant: Shows the target space, the thread and the drafted text, then asks for confirmation before sending.

You: Confirm.

Assistant: Sends the reply under your name in that thread. It does not touch any other message.

Contents

• Quick start
• What you can ask it to do
• How the server acts in Chat
• What can change
• Getting access
• Configuration
• Data, limits and background work
• Technical documentation
• Support

Quick start

You need Node.js 20+ and a Google account with access to Google Chat. Credentials are not required at install time — the server connects from the conversation.

1. Add the server to your AI app.
2. Say "connect Google Chat": the assistant walks you through creating the OAuth client and approving access without editing config files or restarting.
3. Ask the read-only question above.

In the app: open Settings → MCP servers, select Add server, choose STDIO, enter the command npx -y mcp-google-chat@latest, then select Save. No environment variables are needed — connect from the chat afterwards.

From the command line:

codex mcp add google-chat \
  -- npx -y mcp-google-chat@latest

codex mcp list

Codex MCP documentation

claude mcp add \
  --transport stdio --scope user google-chat \
  -- npx -y mcp-google-chat@latest

claude mcp list

Claude Code MCP documentation

The current official path is Settings → Extensions. For a custom desktop extension, open Advanced settings → Extension Developer → Install Extension…, select a .mcpb file and follow the prompts.

This repository currently publishes an npm stdio package and does not contain a .mcpb bundle. For Claude Desktop builds that still support local configuration, use the following JSON stdio configuration as a fallback:

{
  "mcpServers": {
    "google-chat": {
      "command": "npx",
      "args": ["-y", "mcp-google-chat@latest"]
    }
  }
}

In those builds, save it to ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows.

Claude Desktop MCP documentation

Add this to ~/.cursor/mcp.json on macOS/Linux or %USERPROFILE%\.cursor\mcp.json on Windows:

{
  "mcpServers": {
    "google-chat": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "mcp-google-chat@latest"]
    }
  }
}

Cursor MCP documentation

Run MCP: Open User Configuration and add:

{
  "servers": {
    "google-chat": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "mcp-google-chat@latest"]
    }
  }
}

Check it with MCP: List Servers.

VS Code MCP documentation

What you can ask it to do

Catch up on a conversation

• Show my spaces and find the direct message with alex@example.com.
• What was posted in the release space today? Summarize the decisions.
• Show the whole thread this message belongs to.

Post and maintain messages

• Send a status update to the team space.
• Reply in the deploy thread that the rollout is finished.
• Fix the typo in my last message, or delete it entirely.

React and check attachments

• Add a 👍 to the announcement and show who else reacted with what.
• Remove my reaction from that message.
• What files are attached to this message? Show their names and types.

Manage who is in a space

• Who is in this space, and who are its managers?
• Add alex@example.com to the space and make them a manager.
• Remove a former teammate from the space.

How the server acts in Chat

1. With the OAuth refresh credentials the server acts as the signed-in user: messages send under your name, and edits and deletes reach only your own messages and reactions. Membership changes additionally require you to be a space manager.
2. Spaces accept a bare id, but messages, threads, members, reactions and attachments are addressed by the full resource names the API returns — list first, then act on an exact name.
3. A reply targets a thread by its name or by a thread key. By default a send falls back to starting a new thread when the target cannot be threaded; you can ask it to fail instead.
4. Acting as a Chat app — cards, app direct messages, the dedicated attachment endpoint, forced deletes — is a separate Google Cloud configuration; the only bridge here is a service-account access token supplied as GOOGLE_CHAT_ACCESS_TOKEN.

The Chat API cannot search message text — the only message filters are creation time and thread. find_direct_message finds an existing DM but never creates one, and file bytes are not downloaded or uploaded through this server. Space creation and the other uncovered API methods go through raw_request.

What can change

The AI client controls confirmation prompts. The server marks reads, writes and destructive tools so the client can distinguish catching up from posting.

Getting access

Google Chat requires OAuth 2.0; an API key is not enough. There are two ways in, and the first one needs no configuration files.

Connect from the chat (recommended)

Say "connect Google Chat" and the assistant runs the flow with you:

1. setup_instructions prints the checklist: create or select a Google Cloud project, enable the Google Chat API, configure the consent screen and create a Desktop app OAuth client.
2. Download that client's JSON ("Download JSON") and give the assistant its path — set_client stores it owner-only. The secret never goes through the conversation.
3. start_login returns a Google consent link. Open it on this machine and approve; the code comes back to a one-shot listener on 127.0.0.1 (PKCE), never through the chat.
4. finish_login exchanges the code, saves the tokens to ~/.config/mcp-google-chat/credentials.json (mode 0600) and verifies them with a real Chat API call — so a Chat API that is still switched off is caught right there, not on your first real question.

The tokens are re-read on every call, so the connection works immediately — no restart of the AI app. auth_status shows what is connected, logout revokes and deletes it. The login asks for chat.spaces.readonly, chat.messages, chat.messages.reactions and chat.memberships.readonly; the admin scope behind search_spaces is not requested — use the environment path if you need it.

Environment variables (CI, unattended installs)

1. Create or select a Google Cloud project and enable Google Chat API.

2. Configure the OAuth consent screen and create a Desktop app OAuth client.

3. Authorize the Google account you chat as. The OAuth 2.0 Playground can obtain the refresh token when Use your own OAuth credentials is enabled.

4. Request only the scopes your sessions need. For reading and sending, this is enough:

   https://www.googleapis.com/auth/chat.spaces.readonly
   https://www.googleapis.com/auth/chat.messages.readonly
   https://www.googleapis.com/auth/chat.messages.create
   

   The full per-task table — editing and deleting your own messages, reactions, memberships, admin search — is in docs/TOOLS.md.

Testing-mode OAuth refresh tokens can expire after seven days. Publish the OAuth app, or use an Internal app in a Workspace domain, when you need long-lived access. Treat the client secret and refresh token as passwords.

For a quick session, a short-lived token in GOOGLE_CHAT_ACCESS_TOKEN also works — for example from gcloud auth print-access-token with Chat scopes granted. The same variable is how a service-account Chat-app token reaches the server when you need app-only features.

Configuration

Every variable is optional — with none of them the server connects from the chat.

* For the environment path, provide either the OAuth trio or an access token. When set, they win over a stored in-chat login and the server never refreshes or deletes them.

Started without any credentials, the server still completes the MCP handshake; the instructions and the first tool call then name both fixes — the in-chat login (no restart) and the environment variables (restart).

Data, limits and background work

• Requests go to Google Chat. The local server refreshes Google OAuth tokens and calls the Chat API; the token is never sent to any other host. Its anonymous telemetry contains an installation ID, package version, AI client and platform versions, and tool names — never OAuth tokens, message content, tool arguments or prompts. Set ASKADS_TELEMETRY=0 to opt out.
• Google applies per-project and per-user quotas. On 429, the server uses backoff; reads also retry after network and 5xx errors, while writes are never replayed after an uncertain failure — a replayed send would be a duplicate message in a real room. send_message accepts a custom message id that makes a send addressable and deduplicable.
• There is no background polling. The server runs only when called. list_messages can poll a space incrementally by creation time if your AI app supports scheduled tasks; space event subscriptions go through raw_request.

Technical documentation

• MCP capability catalog — task-oriented pages for every tool.
• All tools and inputs
• Development documentation
• Publishing documentation
• Google Chat API reference

Support

Found a bug or need a scenario? Create an issue or write in Telegram.