Integration · Google Ads

Campaign data as MCP queries.

Google Ads runs on the world's largest search inventory. The official MCP makes campaign data accessible via GAQL queries; community alternatives layer in full CRUD.

Google's first-party MCP server (released October 2025) is read-only initially: list_accessible_customers and a single search tool that runs Google Ads Query Language (GAQL) against the customer's account. The agent translates a natural-language prompt into the right GAQL query, runs it, and returns the result. For mutations (create campaigns, update bids, pause ad groups), MCPBundles offers a 24-tool bundle and gomarble-ai offers full CRUD.

Official MCPOAuth + Developer Token, local pipx or Google Cloud Run.

What you can do via MCP

  • “List all my accessible Google Ads accounts.”

    Calls list_accessible_customers via the official MCP and returns the customer IDs and names the agent has access to.

  • “Show me last week's spend by campaign in account 1234567890.”

    Generates a GAQL query against the campaign resource, executes via the search tool, and returns spend grouped by campaign.

  • “Find search terms with high impression share but low conversion rate.”

    Builds a GAQL query joining search_term_view and metrics, filters by impression share and conversion rate thresholds, and returns the underperforming terms.

  • “Pull the QualityScore distribution for campaign Search-Brand-EN.”

    Queries keyword_view filtered by campaign, aggregates by quality_score, and returns the distribution with rationale tags.

  • “Generate a GAQL query that returns top 20 keywords by cost per conversion.”

    Outputs the GAQL syntax with appropriate metrics, segments, and ORDER BY plus LIMIT clauses; the user can run it directly in Google Ads.

Google Ads · MCP

GAQL query via the official MCP. The agent generates the query, executes via search, and returns campaign-level spend and performance.

Google Ads MCP
docs
# Google Ads MCP, called via the agent runtime
mcp.google_ads.search({
  customer_id: "1234567890",
  query: `
    SELECT
      campaign.name,
      metrics.cost_micros,
      metrics.clicks,
      metrics.conversions,
      metrics.average_cpc
    FROM campaign
    WHERE segments.date DURING LAST_7_DAYS
    ORDER BY metrics.cost_micros DESC
  `
})

# Returns campaign rows with spend, clicks, conversions, and CPC.
One command sample showing how the agent talks to Google Ads. The MCP exposes the platform's primitives as tools; the agent translates the prompt into the right call.

MCP integration

Server
pipx run --spec git+https://github.com/googleads/google-ads-mcp.git google-ads-mcp
Auth
OAuth Client ID and Secret + Developer Token + Project ID
Hosting
Local pipx or Google Cloud Run
  • Read-only initially: list_accessible_customers + search via GAQL
  • Community alternatives: MCPBundles 24-tool bundle, gomarble-ai full CRUD
  • GAQL is the universal Google Ads query language; learning it pays off across the platform
  • Cloud Run hosting recommended for team setups; local pipx fine for individuals

Google Ads MCP repo

Visual demonstration

Google Ads· CampaignsLast 7 days
USD

  • Search-Brand-EN

    Active
    Spend$12,840Impr.1.84MClicks42,180CTR2.29%CPC$0.30Conv.1,238

  • Search-Generic-EN

    Active
    Spend$28,160Impr.3.21MClicks61,490CTR1.92%CPC$0.46Conv.1,604

  • PMax-Holiday-2026

    Learning
    Spend$8,420Impr.920KClicks18,930CTR2.06%CPC$0.45Conv.412

  • Display-Remarketing

    Limited
    Spend$3,210Impr.640KClicks4,820CTR0.75%CPC$0.67Conv.94

  • Shopping-Catalog-EU

    Active
    Spend$15,720Impr.2.10MClicks38,640CTR1.84%CPC$0.41Conv.1,082

  • Search-Competitor-Bid

    Paused
    Spend$0Impr.0Clicks0CTR0.00%CPC$0.00Conv.0
Campaign performance returned by GAQL via the MCP. The agent surfaces underperformers and proposes optimization actions; for actual mutations, layer a community MCP with write capability.
Google Ads· Audience2.4M to 3.1M users

Audience name

In-market: Software Buyers (US)

Lookalike source

Customer Match list: paying-customers-2026

  • LocationUnited States
  • Age25 to 54
  • InterestBusiness Software / SaaS Solutions
  • BehaviorRecently visited B2B SaaS websites
  • LookalikeSimilar to converted customers (last 90 days)
Estimated reach72 / 100
An audience definition the agent generated from a brief. Customer Match feeds the lookalike; in-market signals plus demographics narrow the reach to a defensible 2 to 3 million users.

CLI alternative

Google Ads exposes a comprehensive REST API for non-MCP workflows. The gcloud CLI handles authentication and the broader Google Cloud surface. For AI-driven paid-media work the MCP is the primary interface; the REST API fills in for automation jobs and CI integrations the agent runtime is not the right place for.

Pairs with these skills

This integration pairs with the forthcoming paid-media-strategy, ads-performance-analytics skills. The skill landing pages and SKILL.md sources land in subsequent dispatches; cross-link hyperlinks are added when the skill pages ship.

Cost considerations

Google Ads's MCP is free; the platform spend is the cost.

Platform spend, MCP free

The MCP itself is free or included with your platform plan. The dominant cost lives outside the MCP: ad spend on the platform, compute on the warehouse the platform fronts, or whatever the platform's primary unit of value is.

Google's first-party MCP is free. The dominant cost is the ad spend on Google Ads itself; the MCP just shapes how the agent reads or steers the campaigns.

Recommended patterns

  • Treat the agent's API calls as effectively free; the spend the agent triggers (ad spend, platform usage) is the real cost to manage.
  • Set platform-side budget caps before letting an agent push changes; an agent that misreads its prompt can spend real money on the platform itself.
  • Audit logs matter more than usage dashboards here; read-only agents are safer than full-CRUD agents until the workflow is proven.