Imported from reason-machines/mcp-skills (skills/google-meta-ads-ga4-mcp/SKILL.md). Install upstream with npx skills add reason-machines/mcp-skills --skill google-meta-ads-ga4-mcp. Copyright stays with the author.

Google Ads + Meta Ads + GA4 MCP Server

Skill by ara.so — MCP Skills collection.

Overview

The Google Ads + Meta Ads + GA4 MCP server is a unified Model Context Protocol server that provides 250+ tools for managing advertising campaigns and analytics across Google Ads, Meta (Facebook/Instagram) Ads, and Google Analytics 4. It enables AI assistants like Claude, ChatGPT, Cursor, and n8n to perform full read-write operations on these platforms.

Key capabilities:

• 150+ Google Ads tools — Campaign CRUD, keyword research, bidding, ad creation, audiences, extensions, experiments
• 80+ Meta Ads tools — Campaign/ad set/creative management, audience targeting, lead gen, product catalogs
• 20+ GA4 tools — Reporting, audience analysis, property config, attribution, key events
• Cross-platform analytics — Unified ROAS, performance comparison, budget allocation

This is a remote MCP server (hosted), not a local installation. You connect via URL endpoint.

Installation & Setup

Claude Desktop

1. Locate your Claude config file:

   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json

2. Add the server configuration:

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

3. Restart Claude Desktop
4. Look for the MCP connection indicator (🔌) in the bottom-right corner

ChatGPT

1. Navigate to Settings → Connectors → Add custom connector
2. Enter:
   • Name: Google Meta Ads GA4
   • URL: YOUR_MCP_ENDPOINT_URL
3. Click Save
4. The connector will appear in your available tools

Cursor

1. Open ~/.cursor/mcp.json
2. Add:

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

3. Restart Cursor

Windsurf

1. Open ~/.codeium/windsurf/mcp_config.json
2. Add the same configuration as Cursor
3. Restart Windsurf

Claude Code (CLI)

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

n8n

1. Add an MCP Client node to your workflow
2. Configure:
   • Server URL: YOUR_MCP_ENDPOINT_URL
   • Transport: SSE (Server-Sent Events)
   • Authentication: Bearer token (recommended)

Import the pre-built workflow template to get started.

Codex CLI

Add to ~/.codex/config.toml:

[mcp.google-meta-ads-ga4]
url = "YOUR_MCP_ENDPOINT_URL"
transport = "sse"

Authentication & Credentials

This server uses OAuth 2.1 with PKCE for secure authentication. The authentication flow is handled by the hosted MCP server — you don't manage API keys locally.

First-time setup:

1. When you first use a tool, you'll be prompted to authenticate
2. Follow the OAuth flow for the specific platform (Google Ads, Meta, or GA4)
3. Grant the required permissions
4. Credentials are securely stored on the server side

Environment variables (if self-hosting):

• GOOGLE_ADS_DEVELOPER_TOKEN — Google Ads API developer token
• GOOGLE_ADS_CLIENT_ID — OAuth client ID
• GOOGLE_ADS_CLIENT_SECRET — OAuth client secret
• META_APP_ID — Meta app ID
• META_APP_SECRET — Meta app secret
• GA4_PROPERTY_ID — GA4 property ID

Google Ads Tools

List All Campaigns

// Tool: list_campaigns
// Lists all campaigns with performance metrics

const campaigns = await mcp.call_tool("list_campaigns", {
  customer_id: "1234567890",
  include_metrics: true,
  date_range: "LAST_30_DAYS"
});

// Response includes: campaign name, status, budget, impressions, clicks, cost, conversions, ROAS

Create a Search Campaign

// Tool: create_search_campaign
// Creates a new Google Search campaign

const campaign = await mcp.call_tool("create_search_campaign", {
  customer_id: "1234567890",
  campaign_name: "Q2 2026 CRM Software Campaign",
  budget_amount_micros: 50000000, // $50 daily budget (in micros)
  target_cpa_micros: 25000000, // $25 target CPA
  networks: ["SEARCH", "SEARCH_PARTNERS"],
  locations: ["US"], // Country codes
  languages: ["en"],
  start_date: "2026-04-01",
  end_date: "2026-06-30"
});

// Returns: campaign resource name and ID

Add Keywords to Ad Group

// Tool: add_keywords
// Adds keywords with match types and bids

const keywords = await mcp.call_tool("add_keywords", {
  customer_id: "1234567890",
  ad_group_id: "987654321",
  keywords: [
    {
      text: "best crm software",
      match_type: "EXACT",
      cpc_bid_micros: 5000000 // $5 max CPC
    },
    {
      text: "crm for small business",
      match_type: "PHRASE",
      cpc_bid_micros: 3000000
    },
    {
      text: "customer relationship management",
      match_type: "BROAD",
      cpc_bid_micros: 2000000
    }
  ]
});

Keyword Research with Keyword Planner

// Tool: generate_keyword_ideas
// Gets keyword suggestions with volume and CPC data

const ideas = await mcp.call_tool("generate_keyword_ideas", {
  customer_id: "1234567890",
  seed_keywords: ["project management software", "task tracking tool"],
  location_ids: ["2840"], // United States
  language_id: "1000", // English
  include_adult_keywords: false
});

// Returns: keyword text, avg monthly searches, competition, low/high CPC bid

Create Responsive Search Ad

// Tool: create_responsive_search_ad
// Creates an RSA with multiple headlines and descriptions

const ad = await mcp.call_tool("create_responsive_search_ad", {
  customer_id: "1234567890",
  ad_group_id: "987654321",
  headlines: [
    "Best CRM Software 2026",
    "Try Our CRM Free Today",
    "Top-Rated CRM Platform",
    "Manage Customers Easily",
    "CRM Built for Growth"
  ],
  descriptions: [
    "Free 14-day trial. No credit card required. Get started in minutes.",
    "Loved by 10,000+ businesses. Intuitive interface, powerful features."
  ],
  final_urls: ["https://example.com/crm"],
  path1: "crm",
  path2: "free-trial"
});

Get Campaign Performance Report

// Tool: get_campaign_performance
// Retrieves detailed campaign metrics

const performance = await mcp.call_tool("get_campaign_performance", {
  customer_id: "1234567890",
  campaign_id: "111222333",
  date_range: "LAST_7_DAYS",
  metrics: [
    "impressions",
    "clicks",
    "ctr",
    "cost_micros",
    "conversions",
    "conversion_value",
    "cost_per_conversion",
    "roas"
  ],
  segment_by: "date" // Daily breakdown
});

Manage Negative Keywords

// Tool: add_negative_keywords
// Adds negative keywords to prevent unwanted clicks

const negatives = await mcp.call_tool("add_negative_keywords", {
  customer_id: "1234567890",
  campaign_id: "111222333",
  keywords: [
    { text: "free", match_type: "BROAD" },
    { text: "cheap", match_type: "BROAD" },
    { text: "download", match_type: "PHRASE" }
  ]
});

Create Ad Extensions

// Tool: create_sitelinks
// Adds sitelink extensions to campaigns

const sitelinks = await mcp.call_tool("create_sitelinks", {
  customer_id: "1234567890",
  campaign_id: "111222333",
  sitelinks: [
    {
      link_text: "Pricing",
      final_urls: ["https://example.com/pricing"],
      description1: "Flexible plans",
      description2: "Start from $29/mo"
    },
    {
      link_text: "Features",
      final_urls: ["https://example.com/features"],
      description1: "All-in-one CRM",
      description2: "Sales, marketing, support"
    }
  ]
});

Meta Ads Tools

List All Campaigns

// Tool: meta_list_campaigns
// Lists all Meta ad campaigns

const campaigns = await mcp.call_tool("meta_list_campaigns", {
  account_id: "act_1234567890",
  fields: ["name", "status", "objective", "daily_budget", "lifetime_budget"],
  include_insights: true,
  date_preset: "last_30d"
});

Create a Campaign

// Tool: meta_create_campaign
// Creates a new Meta Ads campaign

const campaign = await mcp.call_tool("meta_create_campaign", {
  account_id: "act_1234567890",
  name: "Q2 Lead Generation Campaign",
  objective: "OUTCOME_LEADS", // OUTCOME_AWARENESS, OUTCOME_ENGAGEMENT, OUTCOME_TRAFFIC, OUTCOME_SALES
  status: "PAUSED", // Create paused, enable later
  special_ad_categories: [] // Or ["CREDIT", "EMPLOYMENT", "HOUSING"]
});

Create Ad Set with Targeting

// Tool: meta_create_ad_set
// Creates an ad set with budget, schedule, and targeting

const adSet = await mcp.call_tool("meta_create_ad_set", {
  account_id: "act_1234567890",
  campaign_id: "23850000000000000",
  name: "US Adults 25-45 Desktop",
  optimization_goal: "LEAD_GENERATION",
  billing_event: "IMPRESSIONS",
  bid_amount: 500, // $5.00 in cents
  daily_budget: 5000, // $50.00
  start_time: "2026-04-01T00:00:00+0000",
  end_time: "2026-06-30T23:59:59+0000",
  targeting: {
    geo_locations: {
      countries: ["US"]
    },
    age_min: 25,
    age_max: 45,
    genders: [0], // 0 = all, 1 = male, 2 = female
    publisher_platforms: ["facebook", "instagram"],
    facebook_positions: ["feed", "right_hand_column"],
    device_platforms: ["desktop"]
  }
});

Create Lead Form

// Tool: meta_create_lead_form
// Creates a native lead generation form

const leadForm = await mcp.call_tool("meta_create_lead_form", {
  page_id: "1234567890",
  name: "CRM Demo Request Form",
  follow_up_action_url: "https://example.com/thank-you",
  privacy_policy_url: "https://example.com/privacy",
  questions: [
    { type: "FULL_NAME" },
    { type: "EMAIL" },
    { type: "PHONE_NUMBER" },
    {
      type: "CUSTOM",
      key: "company_size",
      label: "Company Size",
      type: "MULTIPLE_CHOICE",
      options: ["1-10", "11-50", "51-200", "200+"]
    }
  ],
  context_card: {
    title: "Get a Free Demo",
    content: "See how our CRM can help your business grow.",
    button_text: "Get Started"
  }
});

Upload Image Creative

// Tool: meta_upload_image
// Uploads an image for use in ads

const image = await mcp.call_tool("meta_upload_image", {
  account_id: "act_1234567890",
  file_url: "https://example.com/ad-image.jpg",
  // OR
  // file_path: "/path/to/local/image.jpg"
});

// Returns: image hash to use in ad creatives

Create Ad Creative

// Tool: meta_create_ad_creative
// Creates an ad creative with image/video

const creative = await mcp.call_tool("meta_create_ad_creative", {
  account_id: "act_1234567890",
  name: "CRM Demo Ad Creative",
  object_story_spec: {
    page_id: "1234567890",
    link_data: {
      image_hash: "a1b2c3d4e5f6", // From meta_upload_image
      link: "https://example.com/demo",
      message: "Manage your customers better with our all-in-one CRM. Start your free trial today!",
      name: "Try Our CRM Free",
      description: "No credit card required. Get started in minutes.",
      call_to_action: {
        type: "SIGN_UP",
        value: {
          link: "https://example.com/signup"
        }
      }
    }
  }
});

Create Custom Audience

// Tool: meta_create_custom_audience
// Creates a custom audience for targeting

const audience = await mcp.call_tool("meta_create_custom_audience", {
  account_id: "act_1234567890",
  name: "Website Visitors - Last 30 Days",
  subtype: "WEBSITE", // WEBSITE, CUSTOMER_LIST, APP_USERS, ENGAGEMENT
  retention_days: 30,
  pixel_id: "1234567890123456",
  rule: {
    inclusions: {
      operator: "or",
      rules: [
        {
          event_sources: [{
            type: "pixel",
            id: "1234567890123456"
          }],
          retention_seconds: 2592000, // 30 days
          filter: {
            operator: "and",
            filters: [
              {
                field: "url",
                operator: "i_contains",
                value: "/product"
              }
            ]
          }
        }
      ]
    }
  }
});

Create Lookalike Audience

// Tool: meta_create_lookalike_audience
// Creates a lookalike audience from a seed audience

const lookalike = await mcp.call_tool("meta_create_lookalike_audience", {
  account_id: "act_1234567890",
  name: "Lookalike - Top Customers",
  origin_audience_id: "23850000000000000", // Seed custom audience
  lookalike_spec: {
    type: "similarity",
    ratio: 0.01, // Top 1% similarity
    country: "US"
  }
});

Get Ad Performance Insights

// Tool: meta_get_insights
// Retrieves performance metrics with breakdowns

const insights = await mcp.call_tool("meta_get_insights", {
  object_id: "act_1234567890", // Account, campaign, ad set, or ad ID
  level: "ad", // account, campaign, adset, or ad
  date_preset: "last_30d",
  fields: [
    "impressions",
    "clicks",
    "ctr",
    "spend",
    "cpc",
    "cpm",
    "reach",
    "frequency",
    "conversions",
    "cost_per_conversion",
    "roas"
  ],
  breakdowns: ["age", "gender"] // Segment data by demographics
});

Google Analytics 4 Tools

Run Standard Report

// Tool: ga4_run_report
// Runs a standard GA4 report

const report = await mcp.call_tool("ga4_run_report", {
  property_id: "properties/123456789",
  date_ranges: [
    {
      start_date: "30daysAgo",
      end_date: "today"
    }
  ],
  dimensions: [
    { name: "sessionSource" },
    { name: "sessionMedium" },
    { name: "sessionCampaignName" }
  ],
  metrics: [
    { name: "sessions" },
    { name: "totalUsers" },
    { name: "conversions" },
    { name: "totalRevenue" }
  ],
  order_bys: [
    {
      metric: { metric_name: "sessions" },
      desc: true
    }
  ],
  limit: 10
});

Get Realtime Data

// Tool: ga4_run_realtime_report
// Gets current active users and events

const realtime = await mcp.call_tool("ga4_run_realtime_report", {
  property_id: "properties/123456789",
  dimensions: [
    { name: "country" },
    { name: "deviceCategory" }
  ],
  metrics: [
    { name: "activeUsers" },
    { name: "screenPageViews" }
  ],
  minute_ranges: [
    { start_minutes_ago: 5, end_minutes_ago: 0 }
  ]
});

List Key Events (Conversions)

// Tool: ga4_list_key_events
// Lists all conversion events

const keyEvents = await mcp.call_tool("ga4_list_key_events", {
  property_id: "properties/123456789"
});

// Returns: event name, counting_method, custom/default

Get Available Dimensions and Metrics

// Tool: ga4_get_metadata
// Lists all available dimensions and metrics

const metadata = await mcp.call_tool("ga4_get_metadata", {
  property_id: "properties/123456789"
});

// Returns: dimension/metric API names, UI names, descriptions, data types

Run Pivot Report

// Tool: ga4_run_pivot_report
// Creates a pivot table style report

const pivotReport = await mcp.call_tool("ga4_run_pivot_report", {
  property_id: "properties/123456789",
  date_ranges: [
    { start_date: "2026-01-01", end_date: "2026-03-31" }
  ],
  dimensions: [
    { name: "sessionSource" },
    { name: "deviceCategory" }
  ],
  metrics: [
    { name: "sessions" },
    { name: "conversions" }
  ],
  pivots: [
    {
      field_names: ["deviceCategory"],
      limit: 5,
      order_bys: [
        {
          metric: { metric_name: "sessions" },
          desc: true
        }
      ]
    }
  ]
});

Cross-Platform Patterns

Compare Google Ads vs Meta Ads Performance

// Get Google Ads performance
const googleAds = await mcp.call_tool("get_campaign_performance", {
  customer_id: "1234567890",
  date_range: "LAST_30_DAYS",
  metrics: ["cost_micros", "conversions", "conversion_value"]
});

// Get Meta Ads performance
const metaAds = await mcp.call_tool("meta_get_insights", {
  object_id: "act_1234567890",
  level: "account",
  date_preset: "last_30d",
  fields: ["spend", "conversions", "conversion_values"]
});

// Calculate ROAS for each
const googleROAS = googleAds.conversion_value / (googleAds.cost_micros / 1000000);
const metaROAS = metaAds.conversion_values / metaAds.spend;

// Compare and allocate budget

Correlate Ad Spend with GA4 Conversions

// Get ad spend from both platforms
const googleSpend = await mcp.call_tool("get_campaign_performance", {
  customer_id: "1234567890",
  date_range: "LAST_30_DAYS",
  metrics: ["cost_micros"]
});

const metaSpend = await mcp.call_tool("meta_get_insights", {
  object_id: "act_1234567890",
  date_preset: "last_30d",
  fields: ["spend"]
});

// Get GA4 conversion data by source
const ga4Conversions = await mcp.call_tool("ga4_run_report", {
  property_id: "properties/123456789",
  date_ranges: [{ start_date: "30daysAgo", end_date: "today" }],
  dimensions: [{ name: "sessionSource" }],
  metrics: [{ name: "conversions" }, { name: "totalRevenue" }],
  dimension_filter: {
    filter: {
      field_name: "sessionSource",
      in_list_filter: {
        values: ["google", "facebook", "instagram"]
      }
    }
  }
});

// Analyze attribution discrepancies

Unified Campaign Health Dashboard

// Function to check all campaigns across platforms
async function getCampaignHealth() {
  const googleCampaigns = await mcp.call_tool("list_campaigns", {
    customer_id: "1234567890",
    include_metrics: true,
    date_range: "LAST_7_DAYS"
  });
  
  const metaCampaigns = await mcp.call_tool("meta_list_campaigns", {
    account_id: "act_1234567890",
    include_insights: true,
    date_preset: "last_7d"
  });
  
  const ga4Traffic = await mcp.call_tool("ga4_run_report", {
    property_id: "properties/123456789",
    date_ranges: [{ start_date: "7daysAgo", end_date: "today" }],
    dimensions: [{ name: "sessionSource" }],
    metrics: [{ name: "sessions" }, { name: "bounceRate" }, { name: "conversions" }]
  });
  
  // Flag campaigns with ROAS < 2.0, high bounce rate, or declining performance
  const alerts = [];
  
  googleCampaigns.forEach(campaign => {
    if (campaign.roas < 2.0) {
      alerts.push({
        platform: "Google Ads",
        campaign: campaign.name,
        issue: "Low ROAS",
        value: campaign.roas
      });
    }
  });
  
  return { googleCampaigns, metaCampaigns, ga4Traffic, alerts };
}

Common Workflows

Complete Campaign Launch (Google Ads)

1. Create campaign with create_search_campaign
2. Create ad groups with create_ad_group
3. Add keywords with add_keywords
4. Create ads with create_responsive_search_ad
5. Add extensions with create_sitelinks, create_callouts
6. Set up negative keywords with add_negative_keywords
7. Enable campaign with resume_campaign

A/B Test Setup (Meta Ads)

1. Create campaign with meta_create_campaign
2. Create two ad sets with different targeting using meta_create_ad_set
3. Upload images with meta_upload_image
4. Create creatives with meta_create_ad_creative (different copy/images)
5. Create ads linking creatives to ad sets
6. Monitor performance with meta_get_insights and breakdown by ad set

Audience Retargeting Flow

Google Ads:

1. Create remarketing list with create_remarketing_list
2. Add website visitors based on URL rules
3. Apply to campaign with update_campaign_targeting

Meta Ads:

1. Create custom audience with meta_create_custom_audience (website visitors)
2. Create lookalike with meta_create_lookalike_audience
3. Use in ad set targeting

Performance Optimization Routine

// Weekly optimization script
async function optimizeWeeklyCampaigns() {
  // 1. Identify underperforming keywords
  const keywords = await mcp.call_tool("list_keywords", {
    customer_id: "1234567890",
    include_metrics: true,
    date_range: "LAST_7_DAYS"
  });
  
  const lowQualityKeywords = keywords.filter(kw => 
    kw.quality_score < 5 || kw.ctr < 0.02
  );
  
  // 2. Pause low performers
  for (const kw of lowQualityKeywords) {
    await mcp.call_tool("pause_keyword", {
      customer_id: "1234567890",
      keyword_id: kw.id
    });
  }
  
  // 3. Find high-converting search terms
  const searchTerms = await mcp.call_tool("get_search_terms", {
    customer_id: "1234567890",
    date_range: "LAST_30_DAYS"
  });
  
  const winningTerms = searchTerms.filter(st => 
    st.conversions > 5 && st.conversion_rate > 0.05
  );
  
  // 4. Add as exact match keywords
  for (const term of winningTerms) {
    await mcp.call_tool("add_keywords", {
      customer_id: "1234567890",
      ad_group_id: term.ad_group_id,
      keywords: [
        { text: term.text, match_type: "EXACT", cpc_bid_micros: term.avg_cpc * 1.2 }
      ]
    });
  }
  
  // 5. Check budget pacing
  const campaigns = await mcp.call_tool("list_campaigns", {
    customer_id: "1234567890",
    include_metrics: true
  });
  
  for (const campaign of campaigns) {
    const pacingRatio = campaign.spend_last_7d / (campaign.daily_budget * 7);
    
    if (pacingRatio > 0.9 && campaign.roas > 3.0) {
      // Increase budget for high performers
      await mcp.call_tool("update_campaign", {
        customer_id: "1234567890",
        campaign_id: campaign.id,
        daily_budget_micros: campaign.daily_budget * 1.2 * 1000000
      });
    }
  }
}

Troubleshooting

Connection Issues

Symptom: MCP server not appearing in Claude/ChatGPT

Solutions:

1. Verify the URL is correct in your config file
2. Restart the AI assistant completely (quit and relaunch)
3. Check for JSON syntax errors in config (use a JSON validator)
4. Ensure you have network connectivity to the MCP endpoint

Authentication Failures

Symptom: "Authentication required" or "Invalid token" errors

Solutions:

1. Re-run the OAuth flow by using any tool that requires authentication
2. Check that you've granted all required permissions during OAuth
3. Verify your account has appropriate access (MCC structure for Google Ads)
4. For Google Ads: ensure your developer token is approved (not in test mode)
5. For Meta Ads: check that your app has the required permissions (ads_management, ads_read)

API Errors

Symptom: "RESOURCE_EXHAUSTED" or rate limit errors

Solutions:

• Google Ads has per-developer and per-account rate limits
• Meta Ads has per-app rate limits
• Implement exponential backoff in automation scripts
• Batch operations when possible (e.g., add multiple keywords at once)

Symptom: "PERMISSION_DENIED" errors

Solutions:

• Google Ads: Verify you have the correct customer ID (10 digits, no dashes)
• Meta Ads: Ensure the ad account ID includes "act_" prefix
• GA4: Use the full property ID format "properties/123456789"
• Check account access levels (standard vs. admin)

Data Discrepancies

Symptom: GA4 conversions don't match Google Ads conversions

Solutions:

• Check attribution window settings (GA4 vs. Google Ads may differ)
• Verify conversion import settings if using GA4 goals in Google Ads
• Compare date ranges (GA4 uses property timezone, Google Ads uses account timezone)
• Check for conversion tag implementation issues

Tool Not Found

Symptom: AI assistant says it can't find a specific tool

Solutions:

1. Use list_tools() to see all available tools
2. Check the exact tool name (case-sensitive, underscores not hyphens)
3. Verify your MCP server version supports that tool
4. Some tools require platform-specific authentication first

Best Practices

Security

• Never hardcode credentials in scripts — always use environment variables
• Use the OAuth flow provided by the MCP server, don't try to manage tokens manually
• For production automation, implement proper error handling and logging
• Rotate access tokens regularly if self-hosting

Performance

• Batch API calls when possible (add multiple keywords, create multiple ads)
• Use date range filters to reduce data transfer
• Cache frequently accessed data (account hierarchies, metadata)
• Implement rate limiting in automation scripts

Campaign Management

• Always create campaigns in PAUSED status, review before enabling
• Use labels extensively for organization and bulk operations
• Set up conversion tracking before launching campaigns
• Use experiments for A/B testing instead of manual splits

Data Analysis

• Segment data by time period to identify trends
• Use GA4 as source of truth for cross-platform attribution
• Compare platform-reported conversions with GA4 for discrepancies
• Export large datasets to BigQuery for advanced analysis

Additional Resources

• Official Documentation: Model Context Protocol Spec
• Project Repository: [github.com/irinabuht12-oss