Google Ads MCP Server: Connect Your Ad Campaigns to AI Agents
Google Ads is where most B2B and B2C teams spend their performance marketing budget — campaigns, keywords, RSA ads, budgets, search term reports, geographic and device breakdowns. AI agents that can read and manage Google Ads campaigns can research keywords, build ad groups, write copy, analyze performance, and optimize spend — all through natural language.
MCP (Model Context Protocol) gives AI agents structured access to the Google Ads API. There are two ways to connect: Google's official MCP server and MCPBundles' 24-tool campaign management bundle. This guide covers both.
Two Ways to Connect
1. Google's Official MCP Server
Google released an official Google Ads MCP server (googleads/google-ads-mcp) in early 2026. It exposes two tools: list_accessible_customers (list your accounts) and search (run raw GAQL queries). GAQL (Google Ads Query Language) is powerful — you can query any resource, metric, or segment in Google Ads — but you need to know the syntax and resource names.
Pros: First-party, maintained by Google, direct GAQL access for complex queries. Cons: Requires GAQL knowledge, no structured write operations (create campaigns, add keywords, write ads), local Python + OAuth + developer token setup.
On MCPBundles, Google's official server is available as a hosted option — connect with OAuth and use GAQL without managing a local Python process.
2. MCPBundles Campaign Management Bundle — 24 Tools
MCPBundles wraps the full Google Ads API into 24 structured MCP tools organized around real advertising workflows. Your AI creates campaigns, researches keywords, writes RSA ads, manages budgets, analyzes performance by geography and device, and reviews search term reports — all through natural language, no GAQL required.
Pros: Full campaign lifecycle (read + write), keyword research, structured parameters with validation, hosted setup. Cons: REST-based (not Google's official MCP transport).
Both options are available on MCPBundles — use Google's official server for raw GAQL flexibility, the structured bundle for campaign management, or enable both.
What Each Option Covers
| Capability | Google Official (GAQL) | MCPBundles (24 tools) |
|---|---|---|
| List accounts | Yes | Yes |
| Raw GAQL queries | Yes | Yes |
| List/inspect campaigns | GAQL | Dedicated tool |
| Create campaigns | No | Yes (PAUSED default) |
| Update campaigns | No | Yes |
| Remove campaigns | No | Yes |
| Budget management | No | Yes (pacing, updates) |
| Ad group operations | No | Yes (create, list) |
| Keyword management | No | Yes (add, list, negative keywords) |
| RSA ad creation | No | Yes (headlines, descriptions, final URL) |
| Keyword research | No | Yes (ideas + volume/CPC) |
| Search term reports | GAQL | Dedicated tool |
| Geographic performance | GAQL | Dedicated tool |
| Device performance | GAQL | Dedicated tool |
| Schedule performance | GAQL | Dedicated tool |
| Change history | GAQL | Dedicated tool |
| Account summary | No | Yes |
| Setup | Local Python + OAuth + developer token | Hosted, OAuth once |
| Works with | Claude Code, Cursor | Claude, ChatGPT, Cursor, Windsurf, any MCP client |
Google's official server gives you direct access to the full GAQL query surface — anything you can express in GAQL, you can query. The structured bundle trades that flexibility for ease of use: your AI doesn't need to construct GAQL strings, and it gets write operations that the official server doesn't offer.
What AI Agents Can Do With Structured Tools
Research Keywords Before Spending
Two tools that complement the official server's GAQL-only approach:
- Keyword ideas: Give your AI seed keywords or a URL, get back search volume, competition, CPC estimates, and related keyword suggestions. "Find keyword opportunities for our landing page" becomes actionable data.
- Keyword volume: Check exact monthly search volumes, competition index, and top-of-page CPC bids for specific keywords. The same data from Google Ads Keyword Planner — available to your AI in one call.
Build and Manage Campaigns
Full campaign lifecycle from creation to removal:
- Create campaign — Set name, type (Search, Display, Shopping), daily budget, and targeting. Campaigns start PAUSED by default so nothing goes live accidentally.
- Create ad groups — Organize keywords and ads under campaigns with CPC bids.
- Add keywords — Broad, phrase, or exact match. Add negative keywords at campaign or ad group level to control where ads don't show.
- Create RSA ads — 3-15 headlines (30 chars each), 2-4 descriptions (90 chars each). Your AI writes the copy, the tool handles formatting and validation.
- Update budgets — Change daily budget on any campaign.
- Remove campaigns — Permanent removal when a campaign is done (or pause for temporary stops).
Analyze Performance
Seven dedicated analytics tools, each returning structured data your AI can reason about:
| Tool | What it shows |
|---|---|
| Account summary | Spend, clicks, impressions, conversions across all campaigns |
| Campaign performance | Per-campaign metrics over any date range |
| Search terms | What people actually searched to trigger your ads |
| Geographic performance | Performance by country, region, or city |
| Device performance | Desktop vs mobile vs tablet breakdown |
| Schedule performance | Performance by hour and day of week |
| Change history | Who changed what, when — audit trail |
Run Custom GAQL Queries
For anything the structured tools don't cover, the GAQL tool gives your AI the same raw query capability as Google's official server. Use structured tools for common operations, GAQL for edge cases.
Real Workflows
"What keywords should we target for our new product?"
Your AI calls keyword ideas with your product URL, gets back 50 suggestions with volume and CPC, filters to high-volume low-competition terms, and presents a prioritized list. No Keyword Planner UI needed.
"Create a search campaign for our spring sale"
Your AI creates a paused campaign with a $50/day budget, builds an ad group, adds 15 keywords (mix of broad and exact match), writes an RSA ad with 11 headlines and 4 descriptions, adds negative keywords to exclude irrelevant traffic. You review and enable.
"How did our campaigns perform last month by device?"
Your AI pulls device performance for the date range, spots that mobile CPC is 40% higher than desktop with the same conversion rate, and recommends a bid adjustment.
"What are people actually searching to find our ads?"
Search term report for the last 30 days. Your AI identifies wasted spend on irrelevant queries and recommends negative keywords to add.
RSA Ad Best Practices (Built Into the Skill)
Every Google Ads connection on MCPBundles comes with a skill — structured domain knowledge your AI reads before making calls. For RSA ads, it knows:
- 11+ headlines and 4 descriptions for Good+ ad strength
- Zero pins — pinning all positions guarantees Poor strength
- Keywords in 2-3 headlines — match ad group themes
- Diverse remaining headlines — features, benefits, CTAs, social proof (not synonyms)
- Ad strength ≠ performance — Adalysis data on 1M+ ads shows no correlation between strength score and CTR/CPC
Your AI applies these patterns automatically. You describe what the ad should say, it writes copy that follows best practices.
Setup
- Enable the Google Ads bundle on MCPBundles
- Connect via OAuth (one click — your Google account is linked automatically)
- Ask your AI: "List my Google Ads accounts and show me last month's performance"
No developer token to manage, no GAQL to learn, no local Python environment to configure.
FAQ
What's the difference between this and Google's official MCP server?
Google's official server (googleads/google-ads-mcp) provides raw GAQL query access — powerful if you know the syntax. MCPBundles adds 24 structured tools for campaign creation, keyword management, RSA ad writing, and analytics reports. You can use both: the official server for complex GAQL queries and the structured bundle for common operations.
Can my AI create live campaigns that spend money?
Yes, but with a safety default: all campaigns are created in PAUSED status. Your AI builds the campaign structure, you review it in Google Ads UI, and you enable it when ready. Budget updates on existing campaigns take effect immediately.
Does this work with Manager (MCC) accounts?
Yes. If you manage multiple accounts through an MCC, provide the login_customer_id (manager account ID in XXX-XXX-XXXX format). Your AI can then access any child account under the manager.
How does keyword research work without opening Google Ads?
The keyword ideas and keyword volume tools connect directly to Google's Keyword Planner API. Your AI passes seed keywords or a URL and gets back the same data you'd see in the Keyword Planner UI — monthly search volume, competition, CPC estimates, and related suggestions.
Is this read-only?
No. You get full read and write access: create campaigns, ad groups, keywords, and RSA ads; update budgets and campaign settings; remove campaigns. All write operations require the adwords OAuth scope, which is requested during setup.