AI-Powered Google Ads: How MCP Automates Routine Campaign Work.
SpyLead
Against the backdrop of AI and neural networks becoming part of everyday life, more and more tech giants are launching and announcing their own tools for integration with popular AI models. In April 2026, Meta introduced MCP and CLI tools that allow users to manage Meta Ads through a dialogue with an LLM. But many marketers missed the fact that Google responded in kind by releasing its own Google Ads MCP Server. In short, in this traffic source, the era of manually switching between tabs, exporting endless CSV tables, and calmly refreshing dashboards is also gradually changing, if not coming to an end.
Now those who run traffic from Google Ads can connect their ad accounts and agency accounts to AI assistants such as Claude and give commands in chat, without getting stuck in the interface. In this article, we will take a closer look at what this technology is, how it works, and which routine tasks can already be automated today. We will also give a step-by-step guide to connecting MCP for those who want to stop being button-clickers and move into the role of strategists.
What Is Google Ads MCP Server and Why Is It Needed
MCP, or Model Context Protocol, is an open standard developed by Anthropic and announced in fall 2024. With MCP, AI models can connect directly to external services, tools, and platforms. Google Ads was no exception. In simple terms, this is a “plugin” that turns an LLM from an all-knowing theorist into a hands-on assistant. Previously, those who tried to manually connect AI and Google Ads faced two problems at once:
- Manual copy-pasting of data;
- The limited knowledge of AI at the time of training, meaning AI had no idea what was happening in the account yesterday.
As with Meta Ads MCP and CLI, Google Ads MCP Server essentially acts as a bridge that provides a live connection between an AI agent and an ad account. In other words, Claude can “look inside” the account, analyze fresh data, and even make changes at the request of an affiliate marketer.
Already today, through Google Ads MCP, affiliate marketers can delegate tasks such as:
- Instant account audit. Claude or another popular LLM can collect all campaign data in a couple of minutes and provide a prioritized list of what needs to be fixed;
- Generating and testing hundreds of ads. After getting access to the account, AI can analyze landing pages and generate dozens of relevant headlines and descriptions, as well as optimize existing ones;
- Budget control. AI can monitor budgets and send reports;
- Negative keyword research. AI can analyze the search terms report and find queries without conversions.
Step-by-Step Guide to Connecting Google Ads to Claude in 5 Minutes
Google offers two connection paths: a complex one for developers, through a custom server setup, and a simple no-code one through Zapier. We will cover the second option, since it is ideal for affiliate marketers without technical skills.
Getting a Google Ads API Developer Token
This is the only step, but it only needs to be completed once. Moreover, when using integrations through third-party platforms rather than a custom setup, a Developer Token is often not needed at all.
To get this token, go to the API Center inside your Google Ads account and submit an application for a Developer Token. Here you need to specify the company, website, and briefly describe why the token is needed, for example, for internal audit and automation. Usually, the application is approved within 1 to 2 days.
It is also important to understand that Google API Center is available only for manager accounts. If you do not have such an account, you will need to create one.
If you do not have a token yet, you will see a button with wording similar to: “Apply for API access” or “Request developer token.”
For integration with Claude, the description can be written roughly like this:
We plan to use the Google Ads API to automate the management of our own and client campaigns, as well as to create ads, reports, and workflow integrations with artificial intelligence tools and internal marketing systems.
Integration Through Third-Party Platforms, Using Zapier as an Example
Today, at least several workflow automation platforms offer functionality for integrating Google Ads with popular AI assistants. One such platform is Zapier.
All you need is a user account on the platform with a connected Google Ads account. Then, in Claude settings, go to the integrations section and add a new one by selecting the Zapier MCP connector from the list.
For correct setup, after connecting the connector, you can also ask Claude to integrate all available skills by copying the prompt from the MCP tab in your Zapier account.
You can check whether everything works correctly by giving Claude a simple prompt: “Show Google Ads campaign data for the last 7 days.” If Claude returns real numbers, your personal AI assistant is ready to work.
How to Run Your Own MCP Server, the Developer Path
For affiliate marketing agencies, solo affiliate marketers who manage several client accounts through MCC, or those who simply do not want data to pass through third-party services, there is a more advanced path. You can deploy your own MCP server using Google’s official open-source repository. This method requires a basic understanding of working with the terminal and cloud services, but it gives full control, security, and flexibility.
Note. Below is our tested deployment method, which differs from the one described in Google’s official documentation, where the gcloud command-line tool is used. So installing Google Cloud SDK is not required for this path. Authorization is done directly through OAuth Playground. However, the SDK may be useful if you later need to work with other Google Cloud services.
Step 1: Preparing a Project in Google Cloud
For setup, you need a project in Google Cloud, as well as OAuth data. Go to Google Cloud Console using the same Google account from which advertising is run in Google Ads. Then create a new project.
After that, activate two APIs in the APIs & Services → Library tab: Google Ads API and Service Usage API.
Next, configure OAuth Consent Screen. To do this, fill in App Name and email in the Test Users section, or publish the application through Publish App.
After that, create OAuth Credentials. Go back to the APIs & Services tab, go to Credentials, and select Create Credentials → OAuth Client ID → Web application.
In the Authorized redirect URIs field, add:
https://developers.google.com/oauthplayground
It is better to save the Client ID and Client Secret in a separate text document for copying in the next steps.
Step 2: Installing the MCP Server
On Windows, pipx is used for launch. If the OS username in the root folder name contains a space, it is better to set a custom pipx path right away, otherwise errors may appear. To do this, simply copy and paste the following commands into PowerShell:
powershell[Environment]::SetEnvironmentVariable("PIPX_HOME", "C:\pipx", "User")
[Environment]::SetEnvironmentVariable("PIPX_BIN_DIR", "C:\pipx\bin", "User")
Restart PowerShell and move on to deploying the server. Enter the command:
powershellpipx install git+https://github.com/googleads/google-ads-mcp.git
Next, you need to find the path to the binary, because it will be needed in the config:
powershell(Get-Command google-ads-mcp).Source
Step 3: Developer Token and Refresh Token
Take the Developer Token from Google Ads MCC. Admin → API Center. It works with test accounts right away, while for production you will need to apply through the same interface. We already showed the application process in the section on integration through Zapier.
Now generate a Refresh Token in OAuth Playground. To do this, go to OAuth Playground, click the gear icon, and in the dropdown menu check Use your own OAuth credentials. Then insert the Client ID and Secret from Step 1. In the Input your own scopes field, insert:
https://www.googleapis.com/auth/adwords
Click Authorize APIs → log in with the working account → grant access → Exchange authorization code for tokens. Copy the refresh_token field, which starts with 1//.
Step 4: Creating the ADC File
The official server expects Application Default Credentials at the standard path. We create the file manually in PowerShell:
powershell$adc = [ordered]@{
client_id = "YOUR_CLIENT_ID"
client_secret = "YOUR_CLIENT_SECRET"
refresh_token = "YOUR_REFRESH_TOKEN"
type = "authorized_user"
}
$adcPath = "$env:APPDATA\gcloud\application_default_credentials.json"
New-Item -Path (Split-Path $adcPath) -ItemType Directory -Force | Out-Null
$adc | ConvertTo-Json | Set-Content $adcPath -Encoding UTF8
Step 5: Configuring Claude Desktop
Here it is important to know that file locations may differ across Claude Desktop builds. For the MSIX build of Claude Desktop downloaded from the official website, the config is located at:
%LOCALAPPDATA%\Packages\Claude_pzs8sxrjxfjjc\LocalCache\Roaming\Claude\claude_desktop_config.json
Open this file in any editor, for example VS Code, Notepad++, or the standard Notepad, fully replace its contents with the block below, and save it. If there is a Claude-3p folder nearby, you need to duplicate the same file there. In four places, insert your own values: USERNAME, YOUR_PROJECT_ID, YOUR_DEVELOPER_TOKEN, and 1234567890, which is the 10-digit Login Customer ID without hyphens.
json{
"preferences": {
"coworkHipaaRestricted": false,
"coworkWebSearchEnabled": true,
"coworkScheduledTasksEnabled": true,
"ccdScheduledTasksEnabled": true,
"epitaxyPrefs": {
"starred-local-code-sessions": [],
"starred-cowork-spaces": [],
"starred-session-groups": [],
"dframe-local-slice": {
"pinnedOrder": [],
"customGroupAssignments": {},
"customGroupOrder": {}
}
}
},
"coworkUserFilesPath": "C:\\Users\\USERNAME\\Claude",
"mcpServers": {
"google-ads": {
"command": "C:\\pipx\\bin\\google-ads-mcp.exe",
"env": {
"GOOGLE_APPLICATION_CREDENTIALS": "C:\\Users\\USERNAME\\AppData\\Roaming\\gcloud\\application_default_credentials.json",
"GOOGLE_PROJECT_ID": "YOUR_PROJECT_ID",
"GOOGLE_ADS_DEVELOPER_TOKEN": "YOUR_DEVELOPER_TOKEN",
"GOOGLE_ADS_LOGIN_CUSTOMER_ID": "1234567890"
}
}
}
}
Claude Desktop will automatically recreate missing fields in the preferences block on the first launch.
Step 6: Restart and Check
Close Claude Desktop through the system tray and launch it again. The google-ads connector should appear in the connector list.
Now you just need to make sure everything works correctly. To do this, send a simple test prompt in a new Claude chat:
Call google-ads and create a list of available clients.
In response, Claude should provide a list of Customer IDs for the available accounts.
Pitfalls We Faced During Setup
This will help save time:
- The official googleads/google-ads-mcp ignores the GOOGLE_ADS_CLIENT_ID, GOOGLE_ADS_SECRET, and GOOGLE_ADS_REFRESH_TOKEN environment variables. It reads only the ADC file. This detail is clearly not documented;
- Authorization through gcloud auth application-default login can be skipped, because the built-in gcloud OAuth client does not have permission for the adwords scope, and trying to authorize leads to a series of errors. Manually creating the ADC from Web OAuth credentials solves this with one command;
- In the Microsoft Store / MSIX build of Claude Desktop, the config is located inside the package sandbox, not in %APPDATA%\Claude. This was already mentioned above, but it is worth keeping this important detail in mind;
- The refresh token must include the adwords scope. The token from OAuth Playground contains it if the correct scope was specified in Step 3.
Of course, this path is more complex than connecting ready-made connectors through Zapier and other GUI connectors, but it gives full control over what data is processed and how. For individual work and small teams, local installation is the most reasonable compromise between complexity and privacy.
Important! If you still choose the setup scenario through gcloud, keep in mind that if the Google Cloud project has been created but no payment card has been linked for using Artifact Registry, Cloud Build, Cloud Run, and Docker Registry, all subsequent commands in the Terminal will return the FAILED_PRECONDITION error.
Google requires billing to be enabled, even if you may not spend a cent.
Prospects and Possible Downsides
The introduction of MCP servers by giants like Google and Meta is a signal that the industry is moving toward hyperautomation, and the affiliate marketer’s role is changing.
Of course, there are risks. Fully trusting AI with budget management is still dangerous. Before allowing it to make any changes, always check everything manually. AI can misinterpret your prompt. Another nuance is API limits. Google has limits on the number of requests, so you need to be as specific as possible in your prompts.
Conclusion
Integrating Google Ads through MCP is not just another feature, but a powerful tool. For an affiliate marketer, it is an opportunity to get rid of most routine work and focus on what actually brings money: finding winning funnels, creatives, and strategy. Those who master these tools today will be several steps ahead of competitors tomorrow.
Our Telegram Channel - @SpyLeader
For Spy Service please write - @TrafficSponsor