Google Ads MCP Pro
A Model Context Protocol server providing full read and write control over Google Ads, enabling AI assistants to build campaigns, fix targeting, rewrite ads, and manage bidding strategies without opening a browser.
README
<div align="center">
Google Ads MCP Pro
Stop reading dashboards. Start changing them.
A Model Context Protocol server that gives your AI assistant full read and write control over Google Ads — build campaigns, fix targeting leaks, rewrite ads, and switch bidding strategies without ever opening the browser.
</div>
Why this exists
Most Google Ads MCP servers are read-only. They tell you your CPC went up. They can't lower it.
This one closes the loop:
"Impression share is 12%. Why?"
→ queries the API, finds 80% lost to budget
"Fix it."
→ creates a dedicated budget, relinks the campaign,
switches to Maximize Conversions, verifies the change
That whole exchange happens in chat. No browser. No clicking through six settings screens.
Real story: this fork was built while managing a live account. It found a campaign silently locked out of smart bidding by a shared-budget flag, a geo-targeting setting leaking spend to three foreign countries, and a sitelink contradicting the site's own lead flow — then fixed all three through the API. Read-only tooling would have surfaced none of it.
What you can actually do
| Ask your assistant | What happens |
|---|---|
| "Build a search campaign for my new service" | Budget → campaign → ad group → keywords → RSA → sitelinks, in one pass |
| "We're wasting money on competitor searches" | Pulls the search-terms report, adds negative keywords |
| "Only show my ads to people physically in Jordan" | Flips geo targeting to PRESENCE (the leak most accounts never notice) |
| "This ad's strength is Average — improve it" | Rewrites headlines and descriptions in place |
| "Switch to smart bidding" | Handles the shared-budget blocker, then switches strategy |
| "Which niches are worth entering?" | Keyword Planner volumes, competition, and top-of-page CPC |
| "Upload the leads that became paying clients" | Offline conversion upload by GCLID |
Every write tool supports validate_only — a real dry run against Google's API that validates without applying. Ask for a preview before anything touches a live account.
Quick start
1. Requirements
- Python 3.11+
- A Google Ads developer token (apply here)
- OAuth desktop credentials from Google Cloud Console
2. Install
git clone https://github.com/saifshabsug/google-ads-mcp-pro.git
cd google-ads-mcp-pro
pip install -r requirements.txt
3. Configure
cp .env.example .env
GOOGLE_ADS_DEVELOPER_TOKEN=your_developer_token
GOOGLE_ADS_CLIENT_ID=your_oauth_client_id
GOOGLE_ADS_CLIENT_SECRET=your_oauth_client_secret
GOOGLE_ADS_LOGIN_CUSTOMER_ID=1234567890 # your MCC, digits only
GOOGLE_ADS_CREDENTIALS_PATH=./credentials.json
GOOGLE_ADS_AUTH_TYPE=oauth
4. Register with your MCP client
<details> <summary><b>Claude Code</b></summary>
claude mcp add google-ads -- python /full/path/to/google_ads_server.py
</details>
<details> <summary><b>Claude Desktop</b> — <code>claude_desktop_config.json</code></summary>
{
"mcpServers": {
"google-ads": {
"command": "python",
"args": ["/full/path/to/google_ads_server.py"],
"env": {
"GOOGLE_ADS_DEVELOPER_TOKEN": "...",
"GOOGLE_ADS_CLIENT_ID": "...",
"GOOGLE_ADS_CLIENT_SECRET": "...",
"GOOGLE_ADS_LOGIN_CUSTOMER_ID": "...",
"GOOGLE_ADS_CREDENTIALS_PATH": "/full/path/to/credentials.json",
"GOOGLE_ADS_AUTH_TYPE": "oauth"
}
}
}
}
</details>
First run opens a browser for OAuth consent and caches the token.
🧰 The 43 tools
<details open> <summary><b>Read & analyze</b> (12)</summary>
list_accounts · get_account_currency · get_campaign_performance · get_ad_performance · get_ad_creatives · run_gaql · execute_gaql_query · list_resources · get_image_assets · download_image_asset · analyze_image_assets · get_asset_usage
</details>
<details> <summary><b>Build campaigns</b> (6)</summary>
create_campaign_budget · create_search_campaign · create_ad_group · create_responsive_search_ad · add_keywords · add_location_targeting
</details>
<details> <summary><b>Change live settings</b> (11)</summary>
update_campaign_status · update_campaign_name · update_campaign_budget_amount · update_campaign_budget_link · update_campaign_bidding · update_campaign_geo_target_type · update_ad_group_bid · update_ad_status · update_responsive_search_ad · update_keyword_status · set_device_bid_modifier
</details>
<details> <summary><b>Stop waste</b> (2)</summary>
add_negative_keywords · add_ad_group_negative_keywords
</details>
<details> <summary><b>Ad assets</b> (6)</summary>
add_call_asset · add_sitelinks · add_callouts · add_structured_snippet · add_ad_group_sitelinks · add_ad_group_callouts · remove_campaign_asset
</details>
<details> <summary><b>Conversions</b> (3)</summary>
create_conversion_action · update_conversion_action · upload_click_conversion
</details>
<details> <summary><b>Planning</b> (2)</summary>
generate_keyword_ideas · set_ad_schedule
</details>
Hard-won gotchas
Things the API docs bury, discovered by hitting them in production:
| Symptom | Cause | Fix |
|---|---|---|
BIDDING_STRATEGY_TYPE_INCOMPATIBLE_WITH_SHARED_BUDGET |
Smart bidding is rejected on any explicitly_shared budget — even one used by a single campaign |
create_campaign_budget here forces explicitly_shared: false; use update_campaign_budget_link to move an existing campaign |
FIELD_HAS_SUBFIELDS when switching bidding |
Bidding strategies are messages; the field mask must name a subfield | Use maximize_conversions.target_cpa_micros, not maximize_conversions |
| Spend leaking to other countries | Campaigns default to PRESENCE_OR_INTEREST — it targets people merely interested in your region |
update_campaign_geo_target_type → PRESENCE |
CANT_UPDATE_NEGATIVE on a keyword |
It's an ad-group negative, not a positive keyword | Remove and re-add instead |
LAST_3_DAYS rejected in GAQL |
Not a valid date literal | Use BETWEEN 'YYYY-MM-DD' AND 'YYYY-MM-DD' |
⚠️ Safety
This server writes to live advertising accounts that spend real money.
- Start with
validate_only=trueon anything you haven't run before - Test against a low-budget or paused campaign first
credentials.jsonand.envare gitignored — keep it that way- Developer tokens under Test Access only reach test accounts; Basic Access is needed for production
Provided as-is under MIT. You are responsible for what runs against your account.
Credits
Built on the excellent cohnen/mcp-google-ads by Ernesto Cohnen (ixigo), which provided the read/reporting foundation and OAuth handling. This fork keeps all of it and adds 31 write tools, the Google Ads API v24 upgrade, validate_only dry-run support across every mutation, and the production fixes above.
MIT licensed, original copyright preserved — see LICENSE.
Contributing
Issues and PRs welcome. Especially valuable:
- New write tools (audiences, experiments, Performance Max)
- Additional gotchas for the table above
- Tests against sandbox accounts
<div align="center">
Maintained by Saif Shabsug — software consultancy, Amman
If this saved you a browser session, a ⭐ helps others find it.
</div>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.