Provides tools for auditing and managing Google Ads accounts, including campaign and budget planning, building Search and Demand Gen campaigns, setting keywords, ad copy, sitelinks, callouts, negative keywords, location targeting, bidding, performance analysis, and safely previewing, applying, logging, and rolling back changes.
Provides tools for reading GA4 data to verify lead tracking, analyze conversions and performance, and support cross-channel reporting alongside Google Ads and Search Console data.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@google-marketing Audit my Google Ads account for the last 30 days and list the biggest leaks."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Let Claude (or any AI app that supports MCP) work on your Google Ads, GA4, Search Console and Tag Manager, safely.
Ask in plain language. The agent plans campaigns and budgets, builds Search and Demand Gen campaigns, audits accounts, explains what changed and writes client-ready reports. Every change is previewed first, applied only after your yes, logged, and reversible.
You: Audit my Google Ads account for the last 30 days.
Agent: Score 68/100. Biggest leak: 31% of spend is from people outside your target countries
("presence or interest" targeting on 2 campaigns). Fix list below, highest impact first…
You: Fix the location targeting on both.
Agent: Preview (nothing changed yet): 2 campaigns → people IN the location only. Apply?
You: Yes.
Agent: Applied (change 3f9c2a1b). You can undo it any time.Ask | You get |
"Plan our Q4 campaigns" · "How should I split the budget?" | A campaign plan built from the account's own data: which campaigns and types, naming, budget per campaign with a minimum-budget check, bidding stage, goals and negative lists |
"Restructure this account" | A diagnosis of what wastes money today, then the new structure and a transition plan |
Ask | You get |
"Build the Search campaign for " | Themed keywords, ad copy for your approval (character counts, policy-safe wording, second language), then the campaign, ad groups, ads, sitelinks, callouts, goal and negatives, all created paused |
"Launch Demand Gen with these images" | An image check, one ad group per language, an audience from your customer list, lookalikes and custom segments, created paused, then a first-week read |
"Is this campaign ready to launch?" | A pass/fix table (goal, bidding, channels, locations, audience, assets, approvals), then a launch preview |
Plan what to track, build it in Tag Manager from typed recipes, publish after you see the diff, then check what visitors actually get. Every tag, trigger and variable is previewed first; the previous live version is the rollback.
You: Track form leads and WhatsApp clicks on example.com, split business and consumer.
Agent: Plan: generate_lead and contact_click, split by /business/ vs /consumer/, 4 Ads actions, 2 goals,
plus the dataLayer code for your developers. Build it?
You: Yes.
Agent: Preview 1 of 14: CREATE variables "DLV - form_name" (nothing changed). Create it? …
Agent: Workspace vs live version 12: 4 variables, 4 triggers, 6 tags added. Publish as "2026-10-15 lead tracking"?
You: Yes.
Agent: Version 13 is live (rollback: version 12). Live check ✓ every label and event found in gtm.js.
Next: the test matrix, one real lead per row.Ask | Skill | You get |
"Plan lead tracking for example.com" |
| Event matrix, data-layer code for developers, custom dimensions, Ads actions and goals, consent, build order |
"Build the tracking plan" · "Add WhatsApp click tracking" |
| Variables, triggers and tags built and previewed one by one, published on your yes, checked live |
"Ads shows 0 conversions but we got leads" |
| Tag Manager → GA4 → Google Ads checked end to end, fixes handed to |
"Are we double counting?" · "Audit our events" |
| Duplicates, dead tags, unregistered parameters and unused actions, with a fix list |
"Check consent mode" |
| Consent settings on every tag, the Consent Mode v2 default, and a checklist for developers |
Step-by-step guide: docs/tracking.md.
Ask | You get |
"Audit my Google Ads account" | A score out of 100 across measurement, location leaks, wasted search terms, structure, budgets and ad quality, with a prioritised fix list |
"Why did cost per lead go up last week?" | What changed, which campaigns and segments caused it, checked against GA4 leads, and 2–3 actions |
"Weekly check for acme" | Spend vs budget, new search terms to block, disapproved ads, anomalies, with proposed changes as previews |
"Pause the brand campaign" · "Raise the budget to 150" · "Add these negatives" | A preview → your yes → applied, logged, reversible |
Ask | You get |
"Make the September report for acme" | A branded, self-contained HTML report (Ads, GA4, Search Console) you print to PDF |
"Which queries is our blog ranking for?" · "Is the lead form tag firing?" | Search Console and Tag Manager answers from the same login |
More examples: docs/usage.md.
Related MCP server: Google Ads MCP
Read-only until you say otherwise, per account. Agencies can allow changes on one client and keep another read-only.
Preview first. Google validates every change without applying it; nothing happens until you confirm.
Undo. Every applied change is logged with its previous state; ads_rollback puts it back.
Guardrails. New campaigns and ads start paused. Budget jumps above 50% (configurable) need an explicit override. Location targeting is always "people in the location", never Google's default that also includes people merely interested in it.
Your credentials stay on your computer, in a file only you can read. The agent never sees them.
Details: docs/safety.md.
You need Node.js 20+ and a Google account that can open your Google Ads account. Allow about 15 minutes the first time; the full walkthrough with checks is in docs/setup.md.
1. Create your Google Cloud login (once, about 10 minutes). A Cloud project with the Google Ads API enabled, Explorer access, and a Desktop app OAuth client. No developer token needed. → Setup, part 1
2. Sign in and pick your accounts.
npx google-marketing-agent init # opens Google sign-in, lists every account you can reach
npx google-marketing-agent doctor # every line should be ✓3. Add it to your AI app.
App | How |
Claude Code (tools + skills) |
|
Claude Code (tools only) |
|
Claude Desktop, Cursor, others | Add the server below to the app's MCP settings |
{
"mcpServers": {
"google-marketing": { "command": "npx", "args": ["-y", "google-marketing-agent"] }
}
}On Claude Code older than 2.1.275, run /plugin marketplace add omarmfouad25/google-marketing-agent first, then
/plugin install google-marketing-agent. More apps: setup, part 3.
4. Ask. Start with "Run the setup doctor", then "Audit my Google Ads account for the last 30 days."
Part | What it is |
14 skills | Workflows that tell the agent what to check, in which order and what good looks like. Plan: |
63 tools | Google Ads reporting and changes (29), GA4 (11), Search Console (6), Tag Manager (11), accounts and help (6). 37 are always available; the 26 that only change things load when an account allows changes. docs/tools.md |
Knowledge guide | The |
CLI |
|
Skills install with the Claude Code plugin. Other agents that support Agent Skills can load the skills/
folder directly.
Read this | When you want to |
Go from nothing to a working connection, step by step | |
See what to ask for audits, reports, weekly checks and safe changes | |
Plan, build, publish and verify lead tracking in Tag Manager, GA4 and Google Ads | |
Manage several accounts, defaults, report branding and guardrails | |
Understand previews, guardrails, the change log and undo before allowing changes | |
Know exactly what each skill checks and produces | |
Look up every tool and its parameters | |
See how the parts fit together and where your data lives | |
Cost, privacy, developer tokens, which apps work | |
Fix an error |
Does it cost anything? The package is free (MIT). The Google APIs it uses are free; you only pay for your ads.
Do I need a Google Ads developer token? No. Access now comes from your Google Cloud project.
Can it change my campaigns without asking? No. Accounts are read-only by default, and every change needs your confirmation after a preview.
Where does my data go? From Google straight to the AI app on your computer. Nothing is sent anywhere else.
More: docs/faq.md.
Issues and pull requests are welcome. Never paste tokens, customer IDs or client data into an issue. See CONTRIBUTING.md and SECURITY.md.
MIT © Omar Fouad. Not affiliated with or endorsed by Google. Google Ads, Google Analytics, Search Console and Tag Manager are trademarks of Google LLC.
This server cannot be deployed
Google Ads, Reddit Ads, GA4 and Tag Manager in your AI client, with a preview before every change.
Run Google Ads, Meta Ads, GA4 and Search Console from chat: read, audit and launch campaigns.
Read-only Google Search Console and Google Analytics for AI agents.
SEO & marketing toolkit for AI agents: GA4, Search Console, AdSense, GTM, PageSpeed, Trends.
