Google Ads MCP Server
Search & Performance
Connect your AI assistant to Google Ads through one secure MCP to research, build and optimise campaigns — with every change held for your approval.
What is the Google Ads MCP?
The Google Ads MCP Server is the OpenAds integration that connects your AI assistant — ChatGPT, Claude, Claude Code, Cursor or Codex — directly to your Google Ads account through a single secure Model Context Protocol connection. Instead of clicking through the Google Ads interface or exporting spreadsheets, you brief your assistant in plain English and it researches keywords, reads live account data, drafts campaigns and proposes optimisations on your behalf. Google Ads is the highest-intent channel most advertisers run, and it is also the most detailed: match types, Quality Score, bidding strategies, negative lists and search terms all interact. The MCP gives an AI structured, permissioned access to all of it.
What makes this safe to hand to an AI is the separation of duties at the heart of OpenAds. The assistant can freely read and analyse — pulling metrics, search terms and auction insights — but every action that changes your account is held as a proposal until you approve it. Analysis, proposal, approval and execution are distinct steps. You get the speed of an always-on analyst who knows the Google Ads API intimately, without surrendering control of budgets, bids or live campaigns. Nothing spends money or alters a live campaign without an explicit human tap.
Key capabilities
Pull campaign, ad group, keyword and ad metrics across any date range using live Google Ads Query Language (GAQL) queries.
Mine the search terms report to surface converting queries and wasteful matches.
Generate keyword ideas with volume, competition and bid estimates for new campaigns.
Draft Search, Performance Max, Shopping and Display campaigns with ad groups, responsive search ads and extensions.
Propose negative keyword lists to cut irrelevant spend and tighten targeting.
Recommend bid, budget and bidding-strategy changes with pacing and forecast context.
Analyse device, geo, hour-of-day and audience performance to focus budget.
Assemble clear performance reports and export-ready summaries on demand.
What you can do
Analyse performance
Ask for any slice of your account and the assistant runs the underlying GAQL queries for you — comparing campaigns, ad groups, keywords and ads over the periods that matter. It highlights CPA drift, falling conversion rates, wasted impression share and the auction insights behind rising costs, so you understand not just what changed but why.
Research keywords & audiences
Provide a seed term, landing page or product and the assistant returns keyword ideas with search volume, competition and estimated bids, grouped into sensible ad-group themes. It can layer in in-market and custom audiences, and check which existing keywords already cover a theme so you avoid cannibalising your own campaigns.
Find wasted spend
The search terms report is where Google Ads budgets quietly leak. The assistant scans it for high-cost, zero-conversion queries, broad-match spillover and duplicate coverage, then proposes a negative keyword list and match-type tightening. Every removal is shown for approval before it touches the account.
Create campaigns
Describe your objective, offer and budget and the assistant drafts a complete structure — campaign settings, ad groups, keywords, responsive search ads and extensions such as sitelinks, callouts and structured snippets. For retail it can build Performance Max or Shopping listing-group trees. The full draft is presented as a proposal you review before anything goes live.
Optimise & pace budget
The assistant tracks spend against target and flags campaigns pacing over or under budget mid-flight. It proposes bid adjustments, budget reallocation between winners and laggards, and bidding-strategy changes — for example moving a proven campaign to Target CPA — always with the expected impact spelled out.
Generate reports
Ask for a weekly summary, a client-ready performance report or a specific deep dive, and the assistant compiles the numbers, calls out trends and drafts the narrative. Reports are reproducible because they are built from live queries, not stale exports, and can be regenerated in seconds.
Supported campaign types
Campaign typePrimary useTypical workflow SearchCapture high-intent queriesResearch keywords, draft ad groups and responsive search ads, add negatives, launch on approval Performance MaxCross-channel goal-based reachDefine asset groups and audience signals, set budget and target, review draft, approve ShoppingPromote retail productsStructure listing-group tree, set bids by product group, monitor and optimise DisplayAwareness and remarketingSelect audiences and placements, draft responsive display assets, pace budget VideoYouTube reach and considerationSet targeting, refine video ad text, track view and conversion metrics Demand GenVisual, social-style discoveryBuild audience-led asset groups, review creative, launch on approval
Getting started
Sign in to OpenAds and open the integrations directory.
Select the Google Ads MCP Server and choose Connect.
Authorise access to your Google Ads account and grant the requested read and proposal scopes.
Select the specific account or manager (MCC) sub-account you want to work with.
Confirm your approval preferences so write actions are held for your sign-off.
Once connected, point your AI client at the OpenAds MCP endpoint. ChatGPT, Claude, Claude Code, Cursor and Codex all speak the Model Context Protocol, so adding the OpenAds connection makes the full Google Ads toolset available inside the assistant you already use. Brief it in natural language and it handles the API calls behind the scenes.
First prompts to try
Show me the worst-performing keywords by cost with zero conversions in the last 30 days.
Which search terms are draining budget on my Brand campaign? Propose negatives.
Draft a Search campaign for our new pricing page with a ₹2,000 daily budget.
Compare CPA across all campaigns this month versus last month and explain the shift.
Find keyword ideas around "project management software" with volume and bid estimates.
My Performance Max campaign is pacing over budget — what should I change?
Give me a client-ready weekly report for the last seven days.
Which devices and hours convert best? Suggest bid adjustments.
A complete workflow example
Objective: reduce a rising cost per acquisition on a lead-generation Search campaign without cutting volume.
The prompt: "Our Demo Requests campaign CPA has climbed this month. Find out why and propose fixes to bring it back under ₹1,200 without losing leads."
What OpenAds analysed: the assistant pulled 60 days of campaign and keyword metrics, ran the search terms report, and checked auction insights and device performance. It compared the current period against the prior one to isolate what changed.
Findings: three broad-match keywords had begun matching to irrelevant informational queries, accounting for 38% of spend and no conversions; mobile CPA was running 60% above desktop; and one ad group had lost impression share to a new competitor.
Proposed actions: add eleven negative keywords, tighten two broad-match terms to phrase match, apply a −25% mobile bid adjustment, and raise the top ad group's budget by ₹500 per day to recover impression share. Each change was itemised with expected impact.
The approval step: OpenAds presented the bundle as a proposal. You review each line, uncheck anything you disagree with, and approve. Nothing is executed until you do.
The result: on approval, the assistant executes the approved changes through the Google Ads API and confirms each one. It then schedules a follow-up so you can see the CPA trend a week later — with no manual clicking through the interface.
Best practices
Give the assistant your target CPA, ROAS or budget up front so proposals are measured against real goals.
Review the search terms report weekly — it is the single richest source of savings.
Start with analysis and small optimisations before letting the assistant draft whole campaigns.
Read every proposal line by line; approve selectively rather than all at once when stakes are high.
Let bidding strategies gather enough conversion data before asking for major changes.
Safety and control
OpenAds draws a hard line between read and write actions. Reading data — metrics, search terms, keyword ideas, auction insights — happens freely so the assistant can analyse without friction. Any write action that creates a campaign, adds or removes keywords, changes a bid or budget, or pauses and enables entities is held at an approval checkpoint. You see exactly what will change before it happens, approve or reject each item, and retain a record of what was executed. The AI proposes; you dispose. This is what makes it safe to connect a live advertising account to an autonomous assistant.
OpenAds vs manual Google Ads management
DimensionOpenAdsManual management Analysis speedLive queries answered in seconds via natural languageManual navigation, filters and exports Wasted spend detectionSearch terms scanned and negatives proposed automaticallyDepends on periodic manual review Campaign creationFull draft structure generated from a briefBuilt screen by screen by hand ControlEvery write action held for approvalFull control, but every task is manual ReportingReproducible reports on demandAssembled manually from exports
Frequently asked questions
Can the AI spend money or change campaigns without my approval?
No. Reading and analysis are automatic, but any action that creates, edits, pauses or enables something — or changes a bid or budget — is held as a proposal until you approve it. Nothing executes without your sign-off.
Which AI assistants work with the Google Ads MCP?
Any client that supports the Model Context Protocol, including ChatGPT, Claude, Claude Code, Cursor and Codex. You connect once through OpenAds and use the assistant you already prefer.
Does it work with Manager (MCC) accounts?
Yes. During setup you select the specific account or MCC sub-account you want to work with, and you can reconnect to switch between accounts you have permission to access.
What campaign types can it build?
Search, Performance Max, Shopping, Display, Video and Demand Gen, along with the ad groups, responsive search ads, extensions and negative lists that support them.
Is my account data secure?
The connection is authorised through a single secure MCP handshake scoped to the permissions you grant. The assistant only accesses what you allow, and write actions always pass through your approval checkpoint.
Connect it through one secure MCP connection.