Google Ads MCP: query ad accounts and performance

The slow part of a campaign review is pulling numbers: which campaigns spend a lot and convert badly, which keywords to pause. That usually means exporting several Google Ads reports and stitching them together. Google Ads MCP, maintained by the Google Ads API team under the googleads organization, lets an agent query account data through the API and take over that step. The repo had about 1k stars as of 2026-10-06.

What it does

• Account queries: search pulls campaigns, ad groups, keywords and performance metrics using the Google Ads query language.
• Schema before queries: get_resource_metadata lists the queryable fields of a resource such as campaign, and the metrics and segments resources list what’s available, so the model doesn’t guess field names.
• Multiple accounts: list_accessible_customers returns the customer IDs you can reach; accounts under a manager (MCC) work through login_customer_id.
• Tool trimming: tools_config.yaml turns tool groups on or off and changes name prefixes.

Who it’s for

• Paid-search managers running a few accounts who want weekly numbers pulled and odd campaigns flagged.
• Agency analysts who want to query performance across client accounts in plain language.
• Developers building marketing tools who want an agent to write queries against the real schema.

Setup

You need pipx, a Google Cloud project with the Google Ads API enabled, and Application Default Credentials with the adwords scope. A developer token needs at least Explorer access to query production accounts. Client config:

{
  "mcpServers": {
    "google-ads-mcp": {
      "command": "pipx",
      "args": [
        "run",
        "--spec",
        "git+https://github.com/googleads/google-ads-mcp.git",
        "google-ads-mcp"
      ],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "PATH_TO_CREDENTIALS_JSON",
        "GOOGLE_PROJECT_ID": "YOUR_PROJECT_ID",
        "GOOGLE_ADS_DEVELOPER_TOKEN": "YOUR_DEVELOPER_TOKEN"
      }
    }
  }
}

Our take

Community Google Ads servers abound, and some will change bids or create ads. We picked the official one precisely because it’s read-only for now: the agent can look but can’t touch a budget, which makes it a safe way to hand off reporting and analysis. If you want the agent to edit campaigns, you’ll need another tool and the risk that comes with it. The setup is the hurdle: developer-token access levels, an OAuth client and ADC are all on you. The README also notes that your ad data goes to whichever model you connect, and that calls carry an extra header for usage stats. For shared use, the README covers running it as an HTTP service on Podman or Cloud Run. Licensed Apache-2.0.