What does this skill do, and when should you use it?
This skill guides developers through installing and configuring the official open-source Google Ads MCP Server. It covers Python and pipx checks, Google Ads credential setup, and stdio integration with clients such as Claude, Cursor, and Antigravity. It also documents Cloud Run deployment, connection verification, and common troubleshooting steps. Once connected, the server exposes account discovery, resource metadata, and GAQL search tools in a strictly read-only scope.
Checks for Python 3.12+ and pipx; validates the Developer Token, OAuth client credentials, refresh token, and customer IDs; provides pipx install google-ads-mcp and verifies it with google-ads-mcp --help; supplies Claude Desktop MCP JSON and configuration guidance for Cursor, Antigravity, and Cloud Run; recommends /mcp for verification; and explains list_accessible_customers, get_resource_metadata, and search for Google Ads queries.
- A developer wants to query Google Ads campaigns or reporting metrics through natural language.
- A user has the required Google Ads credentials and needs to install the official MCP server.
- A team needs to connect Google Ads to Claude Desktop, Cursor, or Antigravity.
- A developer wants to deploy the read-only MCP server on Google Cloud Run.
How do you install this skill?
- Do not use the Cloud Run --allow-unauthenticated example as-is; add authentication, authorization, least privilege, credential rotation, and auditing first.
- Before execution, verify the current google-ads-mcp release, the intended pipx install versus pipx run workflow, and each client's actual configuration format.
- This is a static-only assessment: installation, authentication, MCP connectivity, and Google Ads queries were not executed; mainland-China reachability and Chinese interaction support are unproven.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
Python 3.12+pipxGoogle Ads API credentials
Install the repository through npx skills add google/skills, then select Google Ads API MCP Server Installation. Prepare the Developer Token, OAuth Client ID, Client Secret, Refresh Token, and Customer ID; if credentials are missing, the source directs users to google-ads-api-quickstart first. Check that python3 --version is 3.12+ and run pipx --version. Then run pipx install google-ads-mcp and verify the installation with google-ads-mcp --help.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/google/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/ads/google-ads-api-mcp-setup" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Guide me through installing and configuring the Google Ads MCP Server for Cursor.
- Retrieve all campaigns from my Google Ads account 1234567890.
Ask a compatible assistant: “Guide me through installing and configuring the Google Ads MCP Server for Cursor.” Configure the client with pipx run google-ads-mcp and the required environment variables, restart the client, then run agy and enter /mcp to verify the server. Test with a query such as “Retrieve all campaigns from my Google Ads account 1234567890.” Customer IDs must contain 10 digits without hyphens.
What are this skill's strengths and limitations?
- Covers the workflow from environment checks through client integration.
- Explicitly directs users to the official `google-ads-mcp` PyPI package.
- Documents stdio, Claude, Cursor, Antigravity, and Cloud Run scenarios.
- Names all three read-only MCP tools and their arguments.
- Requires Python 3.12+, pipx, network access, and complete Google Ads credentials.
- Users without credentials must first complete a separate quickstart skill.
- The source provides no automated test suite or evidence of actual platform testing.
- The server cannot modify bids, pause campaigns, or create advertising assets.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Google Ads MCP Setup Guide this page ✓ Google · Official | 40 · Not recommended | ★ 21k | 12d ago | Apache-2.0 |
| MCP Server Development Guide ✓ Microsoft · Official | 49 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Gemini Enterprise Managed Agents ✓ Google · Official | 40 · Not recommended | ★ 21k | 12d ago | Apache-2.0 |
| Google Data Manager API Audience Import Assistant ✓ Google · Official | 54 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| Data Manager API Event Ingestion Guide ✓ Google · Official | 52 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
Compared with standard Google Ads client libraries, this skill focuses on installing and integrating an MCP Server with AI clients rather than writing direct client-library code. The source also describes local stdio execution and Cloud Run deployment as deployment options.
How did FollowSkills review this skill?
The document recommends environment variables instead of hard-coded credentials and explains stdio transport, no listening port, and read-only scope, providing partial least-exposure and data-flow transparency. However, it recommends Cloud Run with --allow-unauthenticated without access control, rotation, audit, or rollback guidance, and lacks explicit user-confirmation steps, so substantial points are deducted.
The workflow includes prerequisites, environment checks, PATH troubleshooting, and restart guidance, and defaults to the stable PyPI release. However, pipx install and pipx run are not fully reconciled, platform commands are unverified, Cloud Run assumptions are incomplete, and no committed tests are supplied. Under static calibration the score is kept below 10.
The audience, primary scenarios, credential inputs, and several client integrations are described, with a stated read-only boundary. Trigger precision, non-fit cases, Chinese-language support, and mainland-China network reachability are not addressed, while mandatory Antigravity verification may not fit every user, so only a moderate score is justified.
Front matter provides name, compatibility, author, version, and category; the README supplies installation, Apache-2.0 licensing, maintenance language, and issue-reporting paths. Information architecture is readable, but there is no skill-specific changelog, explicit maintenance owner, dependency pinning, FAQ, or adequate known-limitations section, and the supplied tool example is incomplete.
The skill offers reusable commands and configuration examples for local installation, environment setup, stdio integration, and basic troubleshooting, so the core task is plausibly achievable. However, some client behavior is unverified, the Cloud Run example contains security and deployment assumptions, and no statically verifiable representative output is provided; the score remains within the static cap of 7.
Commands, configuration fragments, environment-variable names, and success criteria provide limited auditability. There are no committed tests, CI coverage, pinned dependencies, or third-party execution records; key claims about clients, Antigravity, and server behavior are largely author assertions, so the score is kept below 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →