API

Google Launches Ads MCP Server

Google · Apr 11, 2026

For years, connecting an AI assistant to your Google Ads data meant one of two things: copy-pasting reports into a chat window, or hiring a developer to build custom API integrations from scratch. Google just changed that equation.

The company has officially released an open-source Model Context Protocol (MCP) server for the Google Ads API, now publicly available on GitHub. It's a quiet but significant milestone — one that signals Google's bet on AI agents becoming a standard layer of the modern marketing stack.

What Is MCP, and Why Does It Matter?

MCP (Model Context Protocol) is an open standard, originally developed by Anthropic, that gives large language models a standardized way to securely connect to external tools and data sources. Think of it as a universal adapter: instead of every AI tool needing its own bespoke integration with every platform, MCP creates one common protocol they all speak.

For Google Ads specifically, this means an AI assistant — whether that's Claude, Gemini, or another LLM — can now query your ad account data in plain English, with no custom code required on your end.

How the Google Ads MCP Server Works

The interaction loop is straightforward:

You ask a question — something like "How is my campaign performance this week?"
The LLM discovers available tools — it identifies the Google Ads MCP search capabilities
The MCP server executes the query — running the underlying Python logic against the Google Ads API
Structured results come back — injected directly into the LLM's context window
You get a readable answer — synthesized from live account data By implementing this server, developers eliminate the need to write custom authentication flows, handle API parsing, or manage data formatting. The server handles all of that.

What It Can (and Can't) Do

This initial release is read-only — designed for diagnostics, analytics, and reporting. Here's what's exposed:

list_accessible_customers — Returns all Google Ads customer IDs and account names the authenticated user can access
search — Executes Google Ads Query Language (GAQL) requests to pull campaign metrics, budgets, and status
get_resource_metadata — Retrieves metadata about Google Ads API resource types (helpful for understanding data structure)
discovery-document — Provides the full Google Ads API discovery document so LLMs can understand what's available Some example prompts you can use immediately:

"What customers do I have access to?" "How many active campaigns do I have?" "What's my campaign performance this week?"

What it cannot do — at least in this release — is write back to the platform. Pausing campaigns, adjusting bids, changing budgets, and creating assets all still require the Google Ads UI or a separate API integration. Future iterations may expand beyond read-only, but for now the focus is on insight, not action.

Getting Set Up

The server is Python-based and uses standard OAuth 2.0 or service account authentication. To connect it to any MCP-compatible AI host (like Claude Desktop or Gemini CLI), you add a configuration block to your host's settings.json:

{
  "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/your/credentials.json",
        "GOOGLE_PROJECT_ID": "YOUR_PROJECT_ID",
        "GOOGLE_ADS_DEVELOPER_TOKEN": "YOUR_DEVELOPER_TOKEN"
      }
    }
  }
}

You'll need three things from the Google Ads Developer console before you start:

A developer token (your unique 22-character access string)
A Google Cloud project ID
An authorized credentials.json file (OAuth 2.0 or service account) The server also supports FastMCP's OAuth proxy feature for dynamic user authentication — useful if you're running it as a web service rather than locally.

Why This Is a Big Deal

There are already third-party MCP servers for Google Ads in the market — some with read-write capabilities, some with broader multi-platform coverage (Google + Meta + GA4). But there's something different about an official, Google-maintained implementation.

Deepest API coverage. Because it comes from the platform team itself, the official server has direct, reliable access to the full Google Ads API surface — no intermediary, no translation layer.

Free and open source. Unlike many of the third-party options, Google's implementation costs nothing. For developers and smaller teams who just want to experiment or build something custom, this is the obvious starting point.

A signal of intent. Google shipping an MCP server is an acknowledgment that AI agents querying ad data in natural language is no longer a novelty — it's a workflow pattern worth officially supporting.

Who Should Pay Attention

If you're a developer building marketing tools or internal dashboards, this is a clean, well-documented foundation for AI-powered reporting features. No more writing bespoke Google Ads API authentication from scratch.

If you're a marketer or analyst, the practical value is still somewhat gated behind technical setup — you'll need API credentials and familiarity with MCP configuration. But the direction of travel is clear: natural-language campaign analysis is getting simpler with every release cycle.

If you're an agency, the read-only constraint matters. You can analyze and report, but the actual optimization work stays in the platform UI for now.

What Comes Next

The official documentation explicitly flags that future versions could expand beyond read-only, potentially enabling AI-driven campaign management directly through MCP. That would be a genuinely transformative capability — an AI agent that can not only tell you which ad sets underperformed last week, but immediately act on that information.

For now, the Google Ads MCP server is a solid, free, well-supported foundation. The source code, demos, and examples live at github.com/googleads/google-ads-mcp, and there's an active #ads-api-ai-tools channel on the Google Advertising Community Discord for questions and feature requests.

View original source