Google Ads MCP Server
Enables reading and writing the Google Ads API for full campaign management, with a dry-run/confirm safety flow on every write.
README
Google Ads MCP Server
A local MCP server for reading and writing the Google Ads API. Full campaign management with a dry-run/confirm safety flow on every write.
Prerequisites
- Google Ads Manager (MCC) account with a developer token.
Check its access level: Google Ads UI -> Tools & Settings -> Setup -> API Center.
- Test: can only call test accounts. Apply for Basic access to touch live accounts (approval can take a few days).
- Basic/Standard: ready for live accounts.
- A Google Cloud project with the Google Ads API enabled and an OAuth2 client of type Desktop app (gives you a client ID and client secret).
Setup
python -m venv .venv
. .venv/Scripts/activate # Windows PowerShell: .venv\Scripts\Activate.ps1
pip install -e ".[dev]"
# Mint a refresh token (one-off browser consent):
python scripts/generate_refresh_token.py --client-id XXX --client-secret YYY
cp .env.example .env # then fill in every value
If your OAuth consent screen is still in testing mode, add the Google account you sign in with as a test user first (Cloud Console -> APIs & Services -> OAuth consent screen -> Audience), otherwise consent fails with a verification error.
Once .env holds the client ID and secret, python scripts/remint_refresh_token.py
re-mints an expired or revoked refresh token and rewrites .env in place, without
printing the secret.
Never commit .env; it is gitignored, and it is the only place real credentials
should ever live.
.env keys: GOOGLE_ADS_DEVELOPER_TOKEN, GOOGLE_ADS_CLIENT_ID,
GOOGLE_ADS_CLIENT_SECRET, GOOGLE_ADS_REFRESH_TOKEN,
GOOGLE_ADS_LOGIN_CUSTOMER_ID (your MCC ID, dashes optional, shown
top-right in the Google Ads UI when logged into the manager account).
Run the tests
pip install -e ".[dev]"
pytest -v
Register with Claude Code
Use the venv's absolute python path so registration works regardless of which
directory Claude Code is launched from (the server also resolves .env from the
project root, not the cwd):
# Windows
claude mcp add google-ads -- "<REPO_PATH>\.venv\Scripts\python.exe" -m google_ads_mcp.server
# macOS / Linux
claude mcp add google-ads -- "<REPO_PATH>/.venv/bin/python" -m google_ads_mcp.server
Add --scope user to make the tools available in every project rather than just this one.
Write safety
Every write tool defaults to a preview: it validates the change with the API's
validate_only and returns a confirmation_token. To apply, call the same tool again
with commit=true and that token. The committed operation is byte-for-byte what was
previewed. Nothing spends money until you confirm.
Tools
Read: list_accounts, run_gaql, get_campaigns, get_ad_groups,
get_keywords, get_performance.
Write (preview -> confirm): create_campaign_budget, create_campaign,
update_budget, update_campaign_status, create_ad_group,
update_ad_group_status, create_responsive_search_ad, add_keywords,
add_negative_keywords, apply_mutate.
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。