Google MCP Server
Enables appending formatted content to Google Docs and creating styled Gmail drafts using markdown, with operator approval and optional API key authentication.
README
Google MCP Server
A lightweight Python server that exposes Google Docs and Gmail as HTTP endpoints. Send markdown-formatted content and it renders as properly styled Docs headings/bullets and professional HTML emails — not plain text walls.
Live deployment: https://mcp-server-google-695514226672.europe-west1.run.app
Features
POST /append_to_doc— append formatted content to any Google DocPOST /create_email_draft— create a styled HTML Gmail draft- Markdown rendering: headings, bold, italic, bullet lists, dividers
- Operator approval gate before every Google API call (or auto-approve for pipelines)
- Optional API key authentication via
X-Api-Keyrequest header - Deployed on Google Cloud Run — continuous deployment from GitHub via Cloud Build
Markdown Syntax
Both endpoints accept markdown in their text fields:
| Syntax | Output |
|---|---|
# H1 / ## H2 / ### H3 |
Docs: named heading style · Email: <h1>–<h3> |
- item or * item |
Docs: bulleted list · Email: <ul><li> |
**bold** |
Bold |
*italic* or _italic_ |
Italic |
***bold+italic*** |
Bold + italic |
--- |
Docs: spacing paragraph · Email: <hr> divider |
Quick Start (Local)
1. Google Cloud setup
- Go to Google Cloud Console → APIs & Services → Library
- Enable Google Docs API and Gmail API
- Go to Credentials → Create Credentials → OAuth 2.0 Client ID
- Application type: Desktop app → Download JSON → save as
credentials.jsonin the project root - Add your Google account as a test user under OAuth consent screen → Test users
2. Install and run
python -m venv .venv
.venv\Scripts\activate # Windows
# source .venv/bin/activate # macOS / Linux
pip install -r requirements.txt
cp .env.example .env # edit .env if needed
uvicorn server:app --reload
On first run the browser opens for OAuth. After approving, token.json is saved and reused.
curl http://localhost:8000/health
# → {"status":"ok"}
Environment Variables
| Variable | Default | Description |
|---|---|---|
SERVER_API_KEY |
(unset) | Server-side secret. When set, all endpoints require the HTTP header X-Api-Key: <value> |
APPROVAL_MODE |
terminal |
terminal = operator prompt in TTY; auto = approve all (required on Cloud Run) |
GOOGLE_CREDENTIALS_PATH |
credentials.json |
Path to OAuth client credentials (local dev) |
GOOGLE_TOKEN_PATH |
token.json |
Path to cached OAuth token (local dev) |
GOOGLE_CREDENTIALS_JSON |
(unset) | Raw JSON content of credentials.json — injected by Cloud Run from Secret Manager |
GOOGLE_TOKEN_JSON |
(unset) | Raw JSON content of token.json — injected by Cloud Run from Secret Manager |
PORT |
8000 |
HTTP listen port — Cloud Run injects this automatically |
SERVER_API_KEYvsX-Api-Key:SERVER_API_KEYis the environment variable the server reads.X-Api-Keyis the HTTP request header clients must send. They are not the same thing — do not addX-API-KEYas an env var.
Copy .env.example to .env and fill in values for local development.
API Reference
All mutating endpoints require X-Api-Key when SERVER_API_KEY is set.
Append to a Google Doc
SERVICE_URL="https://mcp-server-google-695514226672.europe-west1.run.app"
API_KEY="PASTE_YOUR_API_KEY_HERE"
cat > payload.json <<'EOF'
{"doc_id":"YOUR_GOOGLE_DOC_ID","content":"# Section\n\n**Key point:** something important.\n\n- Item one\n- Item two"}
EOF
curl -i -X POST "$SERVICE_URL/append_to_doc" \
-H "Content-Type: application/json" \
-H "X-Api-Key: $API_KEY" \
--data-binary @payload.json
{"status": "ok", "doc_id": "YOUR_DOC_ID", "chars_added": 58}
doc_id comes from your Doc URL: https://docs.google.com/document/d/{doc_id}/edit
Create a Gmail draft
SERVICE_URL="https://mcp-server-google-695514226672.europe-west1.run.app"
API_KEY="PASTE_YOUR_API_KEY_HERE"
cat > payload.json <<'EOF'
{"to":"recipient@example.com","subject":"Project Update","body":"# Project Update\n\n**Status:** On track\n\n- Milestone A complete\n- Milestone B in progress\n\n*Regards*"}
EOF
curl -i -X POST "$SERVICE_URL/create_email_draft" \
-H "Content-Type: application/json" \
-H "X-Api-Key: $API_KEY" \
--data-binary @payload.json
{"status": "ok", "draft_id": "r12345abc"}
The draft appears in Gmail → Drafts with full HTML formatting (600 px card, proper heading hierarchy, bullet lists).
Health check (no auth required)
curl https://mcp-server-google-695514226672.europe-west1.run.app/health
# → {"status":"ok"}
Interactive docs: http://localhost:8000/docs (local only)
Docker
# Pre-generate token.json on a machine with a browser (one-time)
python -c "from auth import get_credentials; get_credentials()"
# Build
docker build -t google-mcp-server .
# Run — mount credentials at runtime, never bake them into the image
docker run -d \
--name google-mcp-server \
-p 8000:8000 \
-e SERVER_API_KEY=your-secret-key \
-e APPROVAL_MODE=auto \
-e GOOGLE_CREDENTIALS_PATH=/secrets/credentials.json \
-e GOOGLE_TOKEN_PATH=/secrets/token.json \
-v /path/to/credentials.json:/secrets/credentials.json:ro \
-v /path/to/token.json:/secrets/token.json \
google-mcp-server
Deploy to Cloud Run
The service is live and continuously deployed via Cloud Run's built-in GitHub integration.
| Detail | Value |
|---|---|
| Project | NextLeap (gen-lang-client-0491576843) |
| Service | mcp-server-google |
| Region | europe-west1 |
| URL | https://mcp-server-google-695514226672.europe-west1.run.app |
| Deployment | Cloud Run built-in CD: push to main → Cloud Build → Cloud Run |
| Auth | Allow unauthenticated at Cloud Run level; app enforces X-Api-Key on mutating endpoints |
How it works
- Push to
main→ Cloud Build triggers automatically - Cloud Build builds the Docker image and pushes it to Artifact Registry (managed automatically)
- Cloud Run deploys the new revision
- At startup,
start.shwritesGOOGLE_CREDENTIALS_JSONandGOOGLE_TOKEN_JSON(from Secret Manager) to/tmp, sets path env vars, then starts uvicorn
Cloud Run environment
| Variable | Source | Value |
|---|---|---|
APPROVAL_MODE |
Plain env var | auto |
GOOGLE_CREDENTIALS_JSON |
Secret Manager: google-mcp-credentials:latest |
Raw credentials.json content |
GOOGLE_TOKEN_JSON |
Secret Manager: google-mcp-token:latest |
Raw token.json content |
SERVER_API_KEY |
Secret Manager: google-mcp-api-key:latest |
Backend API key |
For the full setup guide, API key rotation, log commands, and troubleshooting see deployment_plan.md.
Security
| File | Purpose | Committed? |
|---|---|---|
credentials.json |
OAuth 2.0 client secret | No — .gitignore |
token.json |
Cached user OAuth token | No — .gitignore |
.env |
Local secrets | No — .gitignore |
SERVER_API_KEYis stored in Google Secret Manager in production — never in code or committed filesAPPROVAL_MODE=terminalis the safe default locally — requires explicitybefore any Google API call- Cloud Run provides managed TLS on
*.run.appautomatically - No GitHub Actions, no
GCP_SA_KEY/GCP_PROJECT_IDsecrets needed in GitHub
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。