google-mcp-suite-network

google-mcp-suite-network

A tiny TypeScript wrapper that exposes five stdio MCP servers (Gmail, Calendar, Sheets, Docs, Drive) as Streamable HTTP endpoints, enabling a single Docker host to serve all Google services to any MCP client.

Category
访问服务器

README

google-mcp-suite-network

Gmail Calendar Sheets Docs Drive Health

A tiny TypeScript wrapper that takes google-mcp-suite — five stdio MCP servers for Gmail, Calendar, Sheets, Docs, and Drive — and exposes each one over the network as a Streamable HTTP MCP endpoint, so you can run it once on a homelab/Docker host and point any MCP client at it.

MCP client  ──HTTP──►  this proxy  ──stdio──►  google-mcp-<service>  ──►  Google APIs

Each service is mounted at its own path:

Service Endpoint
Gmail /gmail/mcp
Calendar /calendar/mcp
Sheets /sheets/mcp
Docs /docs/mcp
Drive /drive/mcp

Plus GET /healthz for health checks and GET / for a service listing.

Every incoming HTTP session spawns its own child stdio server (identity in this suite is bound per process) and JSON-RPC messages are bridged transparently in both directions.

How it works

  • src/index.ts runs an Express server. On the MCP initialize request it spawns the matching google-mcp-<service> binary and wires its stdin/stdout to a StreamableHTTPServerTransport. The session id maps to that child for follow-up requests; closing either side tears down the other.
  • No tool logic is duplicated — the proxy forwards raw JSON-RPC, so every operation the suite ships is available unchanged.

Prerequisites: Google OAuth

google-mcp-suite needs a Google Cloud OAuth client (Desktop app type) and per-account tokens stored in ~/.google-mcp/. It loads these at startup, so a service endpoint only works once its account is authorized.

  1. In Google Cloud: create a project, enable the Gmail/Calendar/Sheets/Docs/Drive APIs, create a Desktop app OAuth client, and download the client secret.
  2. Save it as client_secret.json.

Authorize an account (recommended: on your workstation)

The consent flow opens a browser and uses a loopback redirect, which is awkward in a headless container. The simplest path is to authorize on a machine with a browser, then ship the resulting ~/.google-mcp/ into the Docker volume.

# On your workstation (Node 22+):
npm i -g google-mcp-suite
mkdir -p ~/.google-mcp && cp client_secret.json ~/.google-mcp/
google-mcp-doctor scopes                 # see required APIs/scopes
google-mcp-doctor auth you@example.com   # browser consent -> writes the token
google-mcp-doctor                        # verify every account is reachable

Then load the populated directory into the named volume used by compose:

docker volume create google-mcp-suite-docker_google-mcp-config
docker run --rm \
  -v google-mcp-suite-docker_google-mcp-config:/dest \
  -v "$HOME/.google-mcp:/src:ro" \
  alpine sh -c 'cp -a /src/. /dest/'

The volume name is <project-dir>_google-mcp-config. Confirm yours with docker volume ls after the first docker compose up.

Alternative: authorize inside the container

docker compose run --rm \
  -v "$PWD/client_secret.json:/home/node/.google-mcp/client_secret.json:ro" \
  google-mcp-suite google-mcp-doctor auth you@example.com

This writes the token into the persistent volume. If no browser is available, the doctor prints a URL to complete consent manually.

Run

cp .env.example .env       # set GOOGLE_MCP_ACCOUNT and (recommended) AUTH_TOKEN
docker compose up -d --build
curl localhost:3000/healthz

Configuration

Variable Default Purpose
PORT 3000 Published port.
GOOGLE_MCP_ACCOUNT — Account label/email bound to every service (must match doctor auth).
<SERVICE>_ACCOUNT — Per-service override, e.g. GMAIL_ACCOUNT, DRIVE_ACCOUNT.
AUTH_TOKEN — If set, every /<service>/mcp request needs Authorization: Bearer <token>.
HOST 0.0.0.0 Bind address.
BODY_LIMIT 50mb Max JSON body (Drive uploads ride inside JSON-RPC).

Connect an MCP client

Point any Streamable-HTTP-capable MCP client at the per-service URL. Example (.mcp.json style):

{
  "mcpServers": {
    "gmail": {
      "type": "http",
      "url": "http://your-homelab-host:3000/gmail/mcp",
      "headers": { "Authorization": "Bearer YOUR_AUTH_TOKEN" }
    },
    "drive": {
      "type": "http",
      "url": "http://your-homelab-host:3000/drive/mcp",
      "headers": { "Authorization": "Bearer YOUR_AUTH_TOKEN" }
    }
  }
}

Drop the headers block if you did not set AUTH_TOKEN.

Local development

npm install
npm run dev      # tsx watch
npm run build    # tsc -> dist/
npm start        # node dist/index.js

Releases & Docker image

Released images are published to Docker Hub at vantreeseba/google-mcp-suite. Pull a pinned version (or latest) instead of building locally:

docker pull vantreeseba/google-mcp-suite:latest

To run the published image, set image: vantreeseba/google-mcp-suite:latest in docker-compose.yml, remove the build: . line, and run docker compose up -d (without --build).

Versioning is automated with semantic-release. On every push to main, GitHub Actions analyzes the Conventional Commits since the last release and, when a release is warranted:

  • bumps the version and updates CHANGELOG.md,
  • creates the Git tag and GitHub release,
  • builds and pushes vantreeseba/google-mcp-suite:<version> and :latest.

Commit messages drive the version bump: fix: → patch, feat: → minor, feat!:/BREAKING CHANGE: → major. Commits like chore:/docs: alone do not trigger a release.

CI setup

The release workflow (.github/workflows/release.yml) requires two repository secrets for Docker Hub auth (Settings → Secrets and variables → Actions):

  • DOCKERHUB_USERNAME — your Docker Hub username (vantreeseba).
  • DOCKERHUB_TOKEN — a Docker Hub access token with Read & Write scope.

GITHUB_TOKEN is provided automatically by Actions.

Security notes

  • Set AUTH_TOKEN whenever the port is reachable beyond localhost. The bearer check is the only access control in front of full read/write access to your Google account.
  • Terminate TLS at a reverse proxy (Caddy / Traefik / nginx) if exposing it beyond your LAN.
  • Tokens live only in the google-mcp-config volume; back it up accordingly.

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