Google Meta Ads GA4

Google Meta Ads GA4 is an MCP server that connects AI assistants to Google Ads, Meta Ads, and Google Analytics 4 through one remote endpoint.

The MCP server is great for paid media teams, agencies, growth marketers, and technical operators who want one chat-based control layer for campaign management, reporting, and cross-platform analysis.

It supports all MCP-compatible client like ChatGPT, Claude Desktop, Claude Code, Cursor, Windsurf, n8n, Codex CLI, Gemini CLI, and Cline.

Connect your ad and analytics accounts once, then use natural-language prompts to inspect performance, create or edit campaigns, work with audiences, run GA4 reports, and compare spend and return across platforms.

Features

• Remote MCP access through a hosted endpoint URL.
• 150+ Google Ads tools for campaigns, ads, keywords, bidding, extensions, analytics, and conversion tracking.
• 80+ Meta Ads tools for campaigns, ad sets, creatives, audiences, lead generation, catalogs, insights, and business assets.
• 20+ GA4 tools for reports, realtime queries, metadata, audiences, attribution settings, property settings, and linked integrations.
• Cross-platform ROAS analysis across Google Ads, Meta Ads, and GA4.
• Side-by-side platform performance comparisons in one chat session.
• Keyword research through Google Ads Keyword Planner data.
• Experiment setup and promotion workflows for Google Ads.
• Lead forms and test leads in Meta Ads.
• Product catalogs and feed operations in Meta Ads.
• GA4 compatibility checks for dimensions and metrics.
• Research workflows through Google search, webpage analysis, image extraction, image search, and landing page analysis.
• OAuth authentication for account connection.
• Stores tokens on the user machine in the described security model.
• Defaults write actions to explicit confirmation.
• Limits access through scoped permissions.

Use Cases

• Pull Google Ads campaign performance for the current month and sort campaigns by ROAS, spend, conversions, or search term quality.
• Create a new Google Search campaign from a plain-English brief that includes budget, targeting, and keyword ideas.
• Pause low-performing campaigns after a review pass in Claude, ChatGPT, or Cursor.
• Add campaign-level negative keywords across multiple Google Ads campaigns after a search term cleanup.
• Generate keyword ideas for new offers through Google Ads volume, CPC, and competition data.
• Compare Google Ads and Meta Ads efficiency for the same date range inside one prompt.
• Audit Meta lead generation campaigns and inspect CPL, creative mix, and audience segments.
• Create Meta lookalike audiences from high-value customer groups.
• Upload Meta image or video assets and attach them to new creatives.
• Review Facebook page posts, identify promotable posts, and turn a post into a boosted campaign.
• Run GA4 real-time reports during launches, promotions, or landing-page tests.
• Compare organic traffic and paid traffic in GA4 across a 30-day or 90-day window.
• Inspect GA4 key events, custom dimensions, and data stream settings before campaign changes.
• Correlate ad spend with GA4 conversion events to find gaps between platform-reported conversions and site-level outcomes.
• Use n8n to trigger repeatable ad reporting or monitoring flows through an MCP client node.
• Use Claude Code or Codex CLI to script repeatable reporting and analysis tasks against live ad accounts.

How To Use It

Prepare the MCP endpoint

Use this JSON structure for clients that read an MCP JSON config file:

{
  "mcpServers": {
    "google-meta-ads-ga4": {
      "url": "YOUR_MCP_ENDPOINT_URL"
    }
  }
}

Use a Claude Code CLI command for terminal registration:

claude mcp add google-meta-ads-ga4 --transport sse YOUR_MCP_ENDPOINT_URL

Use the endpoint URL in n8n through an MCP client node with Streamable HTTP and bearer auth enabled if you want an authenticated automation path.

Add the server to your MCP client

Claude Desktop

Edit claude_desktop_config.json.

macOS path:

~/Library/Application Support/Claude/claude_desktop_config.json

Windows path:

%APPDATA%\Claude\claude_desktop_config.json

Config:

{
  "mcpServers": {
    "google-meta-ads-ga4": {
      "url": "YOUR_MCP_ENDPOINT_URL"
    }
  }
}

ChatGPT

• Open Settings.
• Open Connectors.
• Choose Add custom connector.
• Paste YOUR_MCP_ENDPOINT_URL.
• Set the connector name to Google Meta Ads GA4.
• Save the connector.

Cursor

Edit:

~/.cursor/mcp.json

Add:

{
  "mcpServers": {
    "google-meta-ads-ga4": {
      "url": "YOUR_MCP_ENDPOINT_URL"
    }
  }
}

Windsurf

Edit:

~/.codeium/windsurf/mcp_config.json

Add:

{
  "mcpServers": {
    "google-meta-ads-ga4": {
      "url": "YOUR_MCP_ENDPOINT_URL"
    }
  }
}

n8n

• Create an MCP Client node.
• Set Server URL to YOUR_MCP_ENDPOINT_URL.
• Set transport to Streamable HTTP if your template expects that mode.
• Turn on bearer authentication if your deployment requires it.
• Connect downstream nodes for scheduled reports, alerts, or analysis output.

Claude Code

Run:

claude mcp add google-meta-ads-ga4 --transport sse YOUR_MCP_ENDPOINT_URL

Connect Google and Meta accounts

On first use, you need to authorize your Google and Meta accounts through the MCP connection flow.

Start with read operations

Start with inspection prompts before you touch write actions. Good opening prompts include:

Show me all my Google Ads campaigns and their performance this month
List all active Meta ad campaigns and their spend this week
Show me a realtime report of active users on my website right now
Compare my Google Ads and Meta Ads ROAS for Q1 2026

This first pass helps you confirm account scope, date ranges, and naming conventions.

Move into guided write actions

After you verify account access, move into controlled edits such as pausing campaigns, adding negatives, or creating draft campaign structures.

Examples:

Pause all campaigns with ROAS below 2.0
Add negative keywords "free" and "cheap" to all my Search campaigns
Create a lookalike audience based on my top purchasers
Boost my latest page post with a $100 budget targeting US users

Treat write operations as production changes. Review account IDs, budget amounts, targeting scope, and date windows in every prompt.

Use cross-platform prompts for decisions, not just reports

The most useful part of this server is not basic read access. The real advantage is joint analysis across ad delivery and analytics data.

Examples:

Which platform is driving more conversions per dollar spent?
Show me all campaigns across Google and Meta with spend over $1000 this month
Correlate my GA4 conversion events with Google Ads campaign performance

These prompts fit agency reviews, weekly pacing checks, and channel allocation decisions.

Build automation after manual validation

After your manual prompts return correct results, move common tasks into automation. n8n is the obvious fit for recurring reports, anomaly detection, notification routing, or dashboard prep. Claude Code and Codex CLI fit code-first reporting workflows that save outputs into CSV, HTML, or docs. ([Ryze AI][4])

Apply safe operating rules

• Use read prompts first.
• Use narrow date ranges during early tests.
• Use one account at a time before you query an MCC or business portfolio.
• Use confirmation-heavy prompts for deletions, bulk edits, and budget changes.
• Log important prompt patterns in your internal playbook.
• Keep separate prompts for reporting, diagnosis, and execution.

Available MCP Tools

Google Ads tools

Meta Ads tools

GA4 tools

Research tools

FAQs

Q: Do I need to obtain API keys for Google Ads, Meta Ads, or GA4 before using this server?
A: No. The server uses OAuth 2.1 with PKCE authentication. The first time a tool attempts to access a platform, the client initiates a browser-based authorization flow. You log in with your existing Google or Meta Business account credentials. Tokens store locally on the client machine.

Q: Is this MCP server free to use?
A: Yes. The MCP server endpoint is free. Costs may arise from the AI assistant subscription and from advertising spend managed through the platforms.

Q: Can this server manage Google Ads Manager accounts with multiple child accounts?
A: Yes. The get_account_hierarchy and list_accounts tools expose MCC account structures. Specify the target account ID when calling tools that operate on campaign-level resources.

Q: What campaign types can I create in Google Ads through this server?
A: The server supports creation of Search, Display, Shopping, Performance Max, Demand Gen, Video, and App campaigns. Each campaign type has a dedicated creation tool with type-specific parameter validation.

Q: Can I run custom queries that access metrics not exposed through the named reporting tools?
A: Yes. Google Ads supports custom GAQL queries through the run_gaql_query tool. Meta Ads supports arbitrary Graph API queries through the meta_fetch tool.

Q: Does the server store any of my advertising data or performance metrics?
A: No. The server operates as a stateless proxy. It does not retain request payloads, response data, authentication tokens, or performance metrics.