google-services-mcp
MCP server that gives AI agents direct access to Google services including Gmail, Calendar, Drive, and more, with multi-account OAuth support. Enables reading and sending email, managing calendar events, and interacting with Google files through natural language.
README
google-services-mcp
MCP (Model Context Protocol) server that gives AI agents direct access to your Google services — Gmail, Calendar, Google Meet, Drive, Contacts, Tasks, Sheets, Docs, Slides, YouTube, and Forms — with multi-account OAuth support.
Add one account, wire the server into any MCP-compatible AI client, and your agent can read and send email, manage your calendar, find files in Drive, create documents and spreadsheets, and more — using your real Google data.
Features
- Gmail — send, list, read, modify, and reply to messages
- Calendar — list calendars, create/read/update/delete events, create Google Meet links
- Drive — list, read, upload, update, delete, and share files
- Contacts — list, search, and create contacts
- Tasks — list task lists, create/complete/delete tasks
- Sheets — create spreadsheets, read/write/append cell ranges, batch update
- Docs — create documents, read text, insert/replace text, batch update
- Slides — create presentations, add/delete slides, find-and-replace text
- YouTube — search videos, manage uploads/playlists/subscriptions
- Forms — create forms, add questions, read responses
- Multi-account — connect several Google accounts, set a default, or pass an
accountargument to any tool - OAuth 2.0 — one-time browser authorization; tokens are stored locally and refreshed automatically
60 tools across 11 services, all prefixed
google_. Full reference: docs/TOOLS.md.
Requirements
- Node.js 24+ (CI runs Node 24; local dev on 26)
- A Google Cloud project with the required APIs enabled and an OAuth 2.0 Desktop app client (one-time setup, ~10 minutes — see below)
Install
npm install -g google-services-mcp
Release channels
latest— stable releases, published automatically on every merge tomainbeta— staging builds, published manually frommain(e.g.0.1.1-beta.0)
Try the staging channel without touching your stable install:
npm install -g google-services-mcp@beta
npx -y google-services-mcp@beta --version
All publishes are signed with npm provenance (trusted publishing via GitHub Actions OIDC) — no npm token is stored in CI.
Quick start
-
Add a Google account — opens your browser for one-time authorization:
google-services-mcp add personal -
Register the server with your MCP client (see MCP client configuration for Claude Desktop, opencode, and generic configs).
-
Start using it — ask your agent to "check my email" or "add an event to my calendar". Or add more accounts any time:
google-services-mcp add work.
MCP client configuration
The server speaks MCP over stdio. Point your client at the google-services-mcp
executable and pass your OAuth credentials via environment variables.
opencode
In opencode.json (or your global config):
{
"mcp": {
"google-services-mcp": {
"type": "local",
"command": ["google-services-mcp"],
"enabled": true,
"environment": {
"GOOGLE_MCP_CLIENT_ID": "your-client-id",
"GOOGLE_MCP_CLIENT_SECRET": "your-client-secret"
}
}
}
}
Claude Desktop / Cursor / other MCP clients
In the client's MCP settings file (e.g. claude_desktop_config.json):
{
"mcpServers": {
"google-services-mcp": {
"command": "google-services-mcp",
"env": {
"GOOGLE_MCP_CLIENT_ID": "your-client-id",
"GOOGLE_MCP_CLIENT_SECRET": "your-client-secret"
}
}
}
}
Google Cloud setup (one-time)
Each installation uses its own OAuth client — credentials are per-user and never shared.
1. Create a project
Open the Google Cloud Console, choose
Select a project → New Project, and name it (e.g. google-services-mcp).
2. Configure the OAuth consent screen
APIs & Services → OAuth consent screen:
- User type: External (required for regular Google accounts).
- App name: anything you like (e.g.
Google Service MCP); support email = yours. - Under Audience → Test users, click Add users and enter every Google
account that will use this server. While the app is in Testing mode,
accounts not on this list get
access blockedon the consent screen.
3. Enable the required APIs
APIs & Services → Library, search and Enable each:
Gmail API · Google Calendar API · Google Drive API · People API ·
Google Tasks API · Google Sheets API · Google Docs API ·
Google Slides API · YouTube Data API v3 · Google Forms API
Automated: with the
gcloudCLI installed and authenticated (gcloud auth login), enable all 10 APIs at once — idempotent, safe to re-run:bash scripts/enable-apis.sh --project your-project-id
4. Create OAuth client credentials
APIs & Services → Credentials → Create Credentials → OAuth client ID:
- Application type: Desktop app.
- Authorized redirect URIs:
http://localhost:8787. - Copy the Client ID and Client Secret.
5. Provide the credentials
Set these environment variables (or put them in your MCP client config as above):
GOOGLE_MCP_CLIENT_ID=<your-oauth-client-id>
GOOGLE_MCP_CLIENT_SECRET=<your-oauth-client-secret>
Optional:
GOOGLE_MCP_DIR=~/.google-services-mcp # where config and tokens live (default ~/.google-services-mcp)
GOOGLE_MCP_REDIRECT_PORT=8787 # local loopback port for OAuth (default 8787)
# Full redirect URI override. Takes precedence over REDIRECT_PORT and must
# exactly match a URI registered for this OAuth client in the Cloud Console.
# GOOGLE_MCP_REDIRECT_URI=http://127.0.0.1:8787/oauth2callback
6. Connect your first account
google-services-mcp add personal
A browser opens to the Google consent screen. After you approve, the account is
stored in ~/.google-services-mcp/accounts/ and ready to use. Repeat with a
different name (e.g. work) to connect more accounts.
CLI reference
google-services-mcp # run the MCP server over stdio
google-services-mcp add <name> # add a Google account (opens browser)
google-services-mcp list # list connected accounts
google-services-mcp remove <name> # remove an account
google-services-mcp set-default <name> # set the default account
google-services-mcp status # show config and token health
google-services-mcp --help # show help
For AI agents
- All tools are prefixed
google_, e.g.google_gmail_list,google_calendar_list_events,google_drive_list_files. - The
accountargument selects which connected account a tool uses; omit it to use the default account. - Agents can connect new accounts themselves via the
account_addtool — it opens the browser consent flow without needing the CLI. - See docs/TOOLS.md for the full tool reference.
Security
- OAuth tokens are stored in
~/.google-services-mcp/accounts/<account-name>.jsonwith restrictive file permissions. Keep them private — never commit them. - The server never logs tokens or credentials.
- Scopes requested: Gmail modify, Calendar, Drive, Contacts, Tasks, Sheets,
Docs, Slides, YouTube, Forms, plus
userinfo.emailfor account display. - If a client secret is ever exposed, rotate it in the Google Cloud Console (Credentials → your OAuth client → rotate secret).
Troubleshooting
| Symptom | Cause / fix |
|---|---|
access blocked on consent |
Account not in Test users (setup step 2.3) — add it and retry |
redirect_uri_mismatch |
Redirect URI must be exactly http://localhost:8787 — no trailing slash — and match what your OAuth client allows |
| Tokens stop working after ~7 days | App is in Testing mode — re-run add, or Publish app on the consent screen |
Google hasn't verified this app |
Normal for unverified apps — click Advanced → Continue for personal use |
Local development
git clone https://github.com/nabheet/google-services-mcp.git
cd google-services-mcp
npm install
npm run build # compile to dist/
npm test # vitest, TDD suites
npm run typecheck # tsc --noEmit
npm run dev # run from source with tsx
See AGENTS.md for architecture and conventions.
Releases (maintainers)
All publishing happens from GitHub Actions — no local npm login needed:
- Stable — every merge to
mainauto-publishes the next patch version tolatestwith provenance, then creates avX.Y.Ztag and GitHub release. No manual steps. The next version is derived from the last publishedlateston npm, so repeated merges never collide; an intentional minor/major bump inpackage.jsonis honored. - Staging — manual beta builds from
main: Actions → Publish to npm → Run workflow. Publishes abetabuild (0.1.1-beta.X) under thebetadist-tag with provenance.
Prerelease tags v*-beta* publish to beta with a prerelease GitHub release
(patch/minor builds for milestone testing). Publishing uses npm trusted
publishing (OIDC): configure it once per package at
npmjs.com/package/google-services-mcp/access with the GitHub repository
nabheet/google-services-mcp and workflow name publish.yml.
npm allows one trusted publisher per package and validates the calling
workflow's filename — all channels run from the same publish.yml file.
No NPM_TOKEN secret is required.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.