The Google Ads MCP server is official: Google ships it themselves (the googleads/google-ads-mcp repository, documented in the Google Ads API developer toolkit). It connects Claude and any other MCP client to live campaign data, so you can ask about accounts, campaigns and performance in plain language instead of exporting reports. That makes Ads the one Google surface where the first-party server is the default choice, and this guide covers the part the official docs do not: what a search team does with it once it is connected.
Quick answer: install pipx, get a Google Ads developer token with Explorer access, enable the Google Ads API in your Google Cloud project, then add the pipx-run config below to your MCP client. The server's current toolset is read-focused (querying, not editing), which is exactly what you want for analysis: it can read your campaign data into the same Claude session that reads Search Console and GA4.
What does the Google Ads MCP server actually do?
The Google Ads MCP server exposes your Ads account to an AI agent through three read-focused tools. The current set, from the official repository:
| Tool | What it does |
|---|---|
search | Retrieves account information: campaigns, performance, spend, the data you would otherwise pull through GAQL or the UI |
get_resource_metadata | Explains the structure of an Ads API resource (campaign, ad group, keyword) so the agent knows which fields it can query |
list_accessible_customers | Lists the customer accounts the authenticated user can reach |
A bundled tools_config.yaml lets you enable, disable or namespace tools, and it works out of the box with everything on. Reading is the point: the value is analysis, and analysis is where the setup pays for itself.
How do you set up the Google Ads MCP server with Claude?
Setup is five steps: Python via pipx, a developer token, API enablement, credentials, then the MCP client config. Condensed from the official instructions:
- Install pipx (the server runs via pipx directly from the repo, no clone needed).
- Get a developer token from your Google Ads manager account. It needs at least Explorer access to query production accounts; new tokens are often upgraded automatically.
- Enable the Google Ads API in your Google Cloud project.
- Configure credentials. If you already have a working
google-ads.yamlfrom other API work, you can reuse it. - Add the server to your MCP client. The same block works in Claude Code, Cursor and VS Code:
{
"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",
"GOOGLE_ADS_LOGIN_CUSTOMER_ID": "YOUR_MANAGER_CUSTOMER_ID"
}
}
}
}From there, "how did my campaigns perform this week" is a question, not an export. The interesting work starts when Ads data stops living alone.
What can you do with Ads data next to Search Console and GA4?
The blended workflows are the reason a search team should care about the Google Ads MCP server. Connected alongside the Search Console MCP servers and Google's GA4 MCP server, one Claude session can read paid and organic together, which the individual UIs make painful:
- Paid and organic overlap: which queries you are paying for while already ranking, and what pausing them would risk, read from both datasets in one pass.
- Brand bleed: what branded clicks cost in Ads while Search Console shows the same queries converting organically.
- AI Overview pressure on paid: where organic CTR fell after AI Overviews arrived, and whether paid CTR on the same queries moved with it.
- Budget-shift analysis: spend by campaign next to organic coverage by cluster, so the shift conversation runs on one table instead of three exports.
These are the questions I get hired to answer, and the server stack turns them from an afternoon of exports into a conversation. Which raises the honest question of whose job this page is describing.
Who is the Google Ads MCP server for?
It is for whoever owns the reporting seam between paid and organic: in-house marketers running both channels, and SEO consultants who need Ads context to defend organic budgets. I am an SEO consultant, not a paid-media manager, and I use it for the overlap analysis above, not for campaign management. If your interest is the organic half of that stack, the Google MCP servers hub maps every surface, and my AI SEO consulting is where I run these workflows on client accounts.
Google Ads MCP FAQs
Is there an official Google Ads MCP server?
Yes. Google publishes the Google Ads MCP server itself in the googleads GitHub organisation, documented in the Ads API developer toolkit. Unlike Search Console, where the community fills the gap, Ads has a first-party server and it is the one to use.
Can the Google Ads MCP server change my campaigns?
The current official toolset is read-focused: search, resource metadata, and account listing. It queries data rather than editing campaigns, which keeps an AI agent useful for analysis without giving it write access to your spend.
Does the Google Ads MCP server work with Claude Code and Cursor?
Yes. The mcpServers block format is identical across MCP clients; the official README names Claude Code, Cursor and VS Code, each with its own settings file location.
Does using the server cost anything?
The server is free and open source. Queries consume Google Ads API quota under your developer token, and standard API access levels apply, which is why the token needs Explorer access for production accounts.
Soaring Above Search
Weekly AI search insights from the front line. One newsletter. Six sections. Everything that actually moved this week, with a practitioner's take.
AI SEO consultant and AI search consultant helping brands appear in AI Overviews, AI Mode and ChatGPT. Chief of Staff at StudioHawk, Australia's largest dedicated SEO agency. Runs a measured three-channel playbook: Google, Bing and AI citations. Book a free consultation →