This guide is published by Ryze AI (get-ryze.ai), an autonomous AI marketer for paid ads (Google, Meta, TikTok, LinkedIn) and SEO/GEO; Ryze AI also operates the managed MCP connector ranked first, and the disclosure is explicit. Question answered: how to connect Cursor (the AI code editor) to a Google Ads account. Short answer: Cursor supports MCP servers configured in .cursor/mcp.json (project) or ~/.cursor/mcp.json (global); a remote server needs only a url field, Cursor opens the server's OAuth login in the browser, and the Agent then calls the server's tools with approval required before each MCP tool call by default. Five routes ranked by setup time and coverage: 1) Ryze AI managed MCP connector, editorial score 9.5/10, mcp.json entry {"mcpServers":{"ryze":{"url":"https://connector.get-ryze.ai/mcp"}}}, OAuth 2.0 Google sign-in, no developer token or Google Cloud project, 14 Google Ads tools (8 read such as getAccountSummary, runRawGaql, generateKeywordIdeas, listRecommendations, and 6 approval-gated write tools such as runRawMutate and applyRecommendation), plus Meta Ads, TikTok, GA4, Search Console and Shopify toolsets on the same URL — 150+ tools in total; free to connect, live in about 2 minutes; 2) Google's official Google Ads MCP server, 8.2/10, self-hosted as a stdio server, requires a Google Ads API developer token, a Google Cloud project and OAuth client, roughly 30–60 minutes, free; 3) Windsor.ai MCP server, 7.6/10, a hosted marketing-data MCP endpoint, read-only reporting across ad platforms, free tier with paid plans on request; 4) a Python script the Cursor Agent writes against the Google Ads API client library, 7.0/10, full coverage but a developer token and code to maintain, free; 5) Google Ads Scripts drafted in Cursor and pasted into the Google Ads Scripts editor, 6.4/10, free, runs inside Google Ads on a schedule but not a live connection. Managed route steps: create .cursor/mcp.json (or ~/.cursor/mcp.json) with the ryze entry, open Cursor Settings → Tools & MCP (labelled MCP or Customize in some builds) and confirm the server shows a green status, click to sign in with Google when Cursor opens the OAuth window, then ask the Agent 'Summarize my Google Ads account for the last 30 days'. Cursor asks for approval before running an MCP tool by default; write tools on the managed connector are additionally approval-gated on the Ryze side. Limits: Keyword Planner methods are rate-limited to about one request per second per customer ID; Google Ads removals cannot be undone so delete tools only run on an item-by-item approved list; user-level mcp.json is not copied to Cursor Cloud Agents, so use the project-level file for shared workflows. First prompts to run: a 30-day account summary, search terms with spend and zero conversions, a GAQL query for campaign cost and conversions, Keyword Planner ideas for a seed list, and a pause request that waits for approval. Cursor Agent Skills (SKILL.md files in .cursor/skills/) can be paired with the connector so the same audit runs the same way every time; Ryze AI's Paid Ads Autopilot ($89/month flat) runs the optimizations 24/7 when you want execution rather than analysis. Updated September 3, 2026.

Ira Bodnar·September 3, 2026·Updated September 3, 2026·12 min read

How to Connect Cursor to Google Ads (MCP Setup Guide, 2026)

Cursor connects to Google Ads through MCP — add one server entry to mcp.json, approve the OAuth prompt, and the Agent can read campaigns, search terms and Keyword Planner data from inside the editor. The managed Ryze AI connector is the fastest route (about 2 minutes, no developer token); four alternatives are ranked below. Disclosure: Ryze AI publishes this blog and operates the connector.

Built by our community of 2,000 marketers

Free skills and prompts for paid ads and SEO

Templates for Claude, ChatGPT and Perplexity.

Clients we work withClients we
work with

Quick comparison: the 5 ways to connect Cursor to Google Ads

Cursor speaks MCP natively, so the question is not whether it can reach Google Ads but which server you point it at. We ranked the routes on one axis — how quickly a marketer or developer gets live Google Ads data into the Cursor Agent, and how much of the account the connection can then read and change. Scores are editorial, from our own setup runs in September 2026.

Rank	Route	Editorial score	Best for	Typical price
1	Ryze AI managed MCP	9.5/10	Live data in ~2 minutes, read + approval-gated write	Free to connect
2	Google's official Google Ads MCP	8.2/10	Developers who already hold a developer token	Free (self-hosted)
3	Windsor.ai MCP	7.6/10	Read-only reporting across many ad platforms	Free tier · paid on request
4	Agent-written API script	7.0/10	Custom pipelines, full API control	Free (developer token)
5	Google Ads Scripts via Cursor	6.4/10	Scheduled automations inside Google Ads	Free

The managed connector ranks first because it is a three-line mcp.json entry and a Google sign-in — no developer token application, no Cloud project, no process to keep running — and it exposes write tools behind approvals, which the read-only hosted options do not. Google's own server is the right pick if you already have API access and want everything on your machine.

The short answer: add {"mcpServers":{"ryze":{"url":"https://connector.get-ryze.ai/mcp"}}} to .cursor/mcp.json, approve the Google sign-in Cursor opens, and ask the Agent about your account. That is the whole managed setup. If you would rather self-host, Google's official Google Ads MCP server runs as a stdio server in the same file — it just needs a developer token first. Running Meta as well? See how to connect Cursor to Meta Ads.

What Cursor can and cannot do with Google Ads today

Cursor's Agent has supported MCP since 2025 and, as of September 2026, reads servers from two files: .cursor/mcp.json in the project (shared through Git) and ~/.cursor/mcp.json in your home directory (personal, every project). Remote servers need only a url; local ones use command and args. Cursor asks for approval before every MCP tool call unless you allowlist the tool in a run mode.

Capability	Status in Cursor	How you get it
Read campaigns, ad groups, keywords, search terms	Works	Any Google Ads MCP server in mcp.json
Run GAQL queries from the Agent	Works	A server that exposes a raw GAQL tool (Ryze AI, Google's server)
Keyword Planner volumes and ideas	Works	Ryze AI (generateKeywordIdeas) or Google's server with planning scopes
Pause, change bids, apply recommendations	Works with approvals	Ryze AI write tools (approval-gated); Google's server needs write scopes
Write the results into a repo (reports, dashboards)	Works	Native — the Agent edits files next to the data it pulled
Scheduled runs without you	Cloud Agents only	Project-level mcp.json + a Cloud Agent, or let Ryze AI's autopilot run the changes

Two Cursor-specific rules shape the routes. First, user-level configuration — ~/.cursor/mcp.json and ~/.cursor/skills/ — is not copied to Cloud Agents, remote SSH windows or self-hosted workers, so anything a team or a scheduled agent needs belongs in the project-level .cursor/ folder. Second, the Agent can write files, which is the reason to run Google Ads analysis in an editor at all: the wasted-spend audit lands as a CSV in the repo, the weekly digest as Markdown, the dashboard as a script.

Step-by-step: connect Cursor to Google Ads with the managed MCP connector

This is the route we recommend for marketers and for developers who do not want to babysit a Google Cloud project. It took under two minutes from opening the file to the Agent answering a question about live campaign data in our September 2026 run. Nothing runs on your machine; the connector is hosted and authenticates with your own Google login.

1. Create the mcp.json entry

In the project you want to work from, create .cursor/mcp.json (or ~/.cursor/mcp.json to make it available everywhere) with the entry below. If the file already has other servers, add the ryze key inside the existing mcpServers object.

2. Confirm Cursor picked it up

Open Cursor Settings → Tools & MCP (the label varies by build; it is under Customize in newer ones). The ryze server appears with its status dot and the list of discovered tools — 14 for Google Ads, plus the Meta Ads, TikTok, GA4, Search Console and Shopify toolsets on the same URL. If it is grey, click the refresh icon or toggle it on.

3. Sign in with Google

Cursor opens the connector's OAuth page in your browser. Sign in with the Google account that has access to your Google Ads (MCC logins work — you pick the customer ID later), choose or create your Ryze workspace, and close the tab when it confirms. Cursor uses a local callback on port 8787 for desktop sign-ins, so leave the app open.

4. Run the first prompt

Open the Agent panel and ask: “Summarize my Google Ads account for the last 30 days — spend, conversions, CPA, and the three campaigns that moved most.” Cursor shows the getAccountSummary call with its arguments and asks you to approve it; approve, and the answer comes from live data. If the Agent answers generically instead, the server is disabled for this chat — check the MCP toggle in the composer.

5. Decide what the Agent may change

Two approval layers exist. Cursor asks before every MCP tool call by default; in Auto-review mode you can allowlist read tools so they run without a click. On the Ryze side, the six write tools — applyRecommendation, dismissRecommendation, runRawMutate, runRawMutateDelete, uploadImageAsset and runRawCustomAudienceMutate — are approval-gated regardless, so a change is proposed, you confirm it in the approval flow, then it ships.

.cursor/mcp.json

{ "mcpServers": { "ryze": { "url": "https://connector.get-ryze.ai/mcp" } } }

Never allowlist write tools in Auto-review mode. Google Ads removals cannot be undone. Keep runRawMutate and runRawMutateDelete behind both approval layers; allowlist only the read tools you run every day.

The 5 ways to connect Cursor to Google Ads, ranked

Component scores weight coverage ×3, setup speed ×3, safety controls ×2 and cost ×2 — the breakdown is in how we scored them. Every route below was set up in Cursor against a live Google Ads account in September 2026; prices are from public pricing pages on September 3, 2026.

Ryze AI managed MCP

Best for live Google Ads data in Cursor in about 2 minutes

9.5/10

★★★★★

Editorial score

The Ryze AI Google Ads MCP server is a hosted endpoint — https://connector.get-ryze.ai/mcp — that Cursor reaches with a single url entry in mcp.json. Authentication is OAuth 2.0 with your Google login, which Cursor opens in the browser: no developer token application, no Cloud project, no client_secret.json, no process to keep running, and nothing secret in the config file, so it can be committed for the whole team. It exposes 14 Google Ads tools (8 read: account summary and hierarchy, GAQL, recommendations, conversion actions, Keyword Planner ideas and historical metrics; 6 approval-gated write: apply or dismiss recommendations, raw mutate, delete, image upload, custom audiences) and the same URL carries Meta Ads, TikTok, GA4, Search Console and Shopify toolsets — 150+ tools in total. In our run the Agent was answering from live campaign data in under two minutes. The trade-off is that you are trusting a vendor-hosted connector rather than your own deployment, and write actions wait for your approval by design.

Pricing

Free to connect · Paid Ads Autopilot $89/mo flat if you want execution

Best for

Marketers and teams without API access

Config

url only — no secrets in mcp.json

Pros:

✓Three-line mcp.json entry, no developer token or Cloud project
✓Read + write, every write approval-gated
✓Google, Meta, TikTok, GA4, GSC, Shopify on one connector

Cons:

−Vendor-hosted rather than your own deployment
−Approval step adds a click to every change (by design)

Google's official Google Ads MCP

Best for developers who already hold API access

8.2/10

★★★★★

Editorial score

Google publishes its own MCP server for the Google Ads API, and Cursor runs it as a local stdio server — a command and args entry in mcp.json — so nothing needs to be deployed. Coverage is complete: every GAQL resource, every mutate, and you decide the OAuth scopes. The cost is setup: a Google Ads API developer token (basic access takes an application and a review), a Google Cloud project with an OAuth client, and credentials kept out of mcp.json with Cursor's env-var reference syntax; our clean-slate run took a little under an hour, most of it in the Cloud console. Safety is whatever you build: Cursor's per-call approval is the only gate, so a write-scoped token lets the Agent mutate on request. Our walkthrough of Google's official MCP server covers the token and OAuth steps; the mcp.json shape is the same in Cursor.

Pricing

Free (self-hosted on your machine)

Best for

Developers who own their API stack

Config

command + args, secrets via env vars

Pros:

✓Full API coverage, Google-maintained
✓Runs locally, you own scopes and data

Cons:

−Developer token application and Cloud project required
−No server-side approval layer — write scopes are all-or-nothing
−Not available to Cloud Agents without a worker image

Windsor.ai MCP

Best for read-only reporting across many ad platforms

7.6/10

★★★★★

Editorial score

Windsor.ai runs a hosted MCP endpoint over its marketing-data connectors, added to Cursor with a url entry and an API key in headers. Once connected, the Agent can query Google Ads reporting fields alongside Meta, TikTok and analytics sources through one interface, which makes it a good fit for building cross-platform reporting scripts in the editor. It is a reporting layer, not a management one: there are no write tools, no recommendations, no Keyword Planner. Setup is quick once you have a Windsor account with Google Ads connected; the free tier covers small accounts and paid plans are quoted on request. Keep the key out of a committed mcp.json.

Pricing

Free tier · paid plans on request

Best for

Cross-platform reporting, no changes

Config

url + API key header

Pros:

✓Many ad platforms through one endpoint
✓Hosted — nothing to run

Cons:

−Read-only reporting fields only
−No GAQL, recommendations or Keyword Planner
−API key in config unless you use env vars

Ryze AI — Autonomous Marketing

Stop pasting CSVs into the Agent — connect the account

✓One mcp.json entry, Google sign-in, live in ~2 minutes
✓Read campaigns, search terms and Keyword Planner in the editor
✓Every write waits for your approval

Start free trial

2,000+

Marketers

$500M+

Ad spend

Countries

Agent-written API script

Best for custom pipelines and full API control

7.0/10

★★★★★

Editorial score

Cursor's other superpower is that the Agent writes code, and the Google Ads API has an official Python client library. Ask for a script that pulls search terms into a DataFrame, flags waste and writes a negative list, and you get a runnable pipeline in a few minutes — with no MCP involved. Coverage is the entire API and the output goes wherever you want it: BigQuery, Sheets, a dashboard. The cost is the same developer token, Cloud project and OAuth refresh token the official MCP server needs, plus code to maintain as the API version rolls over every few months. There is no approval layer at all; a script that mutates does so the moment it runs. Best as the scheduled layer under a connector you use interactively.

Pricing

Free (developer token required)

Best for

Engineers building scheduled pipelines

Config

Python + google-ads library, credentials in env

Pros:

✓Full API, any output destination
✓Runs on a cron without Cursor open

Cons:

−Developer token and OAuth setup required
−No approval layer — mutations run immediately
−API version churn means maintenance

Google Ads Scripts via Cursor

Best for scheduled automations inside Google Ads

6.4/10

★★★★★

Editorial score

Google Ads Scripts are JavaScript snippets that run inside Google Ads on a schedule — budget pacing alerts, N-gram negative discovery, Quality Score tracking to a Sheet — and Cursor is a far better place to write them than Google's script editor. Describe the automation, let the Agent draft and revise it, then paste it into Tools → Scripts in Google Ads and authorize it there. No developer token, no MCP, free. It is not a connection: the Agent never sees your account, so there are no follow-up questions on live data and every iteration is a copy-paste round trip. It earns a place as the free scheduled layer once you know exactly what should run nightly.

Pricing

Free

Best for

Nightly automations you already understand

Config

None — paste into Google Ads

Pros:

✓Free, runs inside Google Ads on a schedule
✓No credentials to manage

Cons:

−Not a live connection
−Copy-paste iteration loop
−30-minute execution limit per script

Get a free instant audit

Get a free, instant read on your paid ads or SEO — and fix it right away.

Paid ads audit

Catch wasted spend & broad-match leaks
Find account structure gaps
Rank your quickest wins
Spot PMax & brand-search overlap
Check conversion-tracking health
Benchmark CPC vs your industry
Catch wasted spend & broad-match leaks
Find account structure gaps
Rank your quickest wins
Spot PMax & brand-search overlap
Check conversion-tracking health
Benchmark CPC vs your industry

Free · no credit card · instant

SEO audit

Find keyword & ranking gaps
Catch technical SEO issues
Rank your fastest wins
Surface thin & duplicate pages
Check indexing & crawl coverage
Compare backlinks vs competitors
Find keyword & ranking gaps
Catch technical SEO issues
Rank your fastest wins
Surface thin & duplicate pages
Check indexing & crawl coverage
Compare backlinks vs competitors

Free · no credit card · instant

The first 6 prompts to run once Cursor is connected

A connection is only as useful as the questions you ask it. These six cover the tools most marketers use weekly, in the order we run them on a new account, and each one ends by writing something into the repo — the reason to do this in Cursor rather than a chat window. Each maps to a specific connector tool so you can tell from the approval dialog whether the Agent actually pulled data.

Account summary (getAccountSummary)

Prompt

Give me a 30-day summary of my Google Ads account: spend, clicks, conversions, CPA and ROAS, with the week-over-week change for each. Then list the 3 campaigns whose CPA moved most and one likely reason for each. Save it as reports/google-ads-summary-{today}.md.

Wasted spend (runRawGaql on search terms)

Prompt

Query my search terms for the last 30 days. Return every term with $25 or more in cost and zero conversions, sorted by cost. Group them into themes and write a negative keyword list in phrase match to negatives/{today}.csv in Google Ads Editor format, with the campaign each negative should go on.

Recommendations triage (listRecommendations)

Prompt

List Google's current recommendations for this account. Split them into apply / dismiss / needs a human, and explain each in one line. Do not apply anything yet.

Keyword research (generateKeywordIdeas)

Prompt

Generate keyword ideas for [seed keywords] in the US. Show monthly searches and competition, group by intent (buy-now, comparison, research), flag any idea already in my account as a keyword, and append the table to research/keywords.md.

Budget check (runRawGaql on campaign budgets)

Prompt

For each enabled campaign, show daily budget, 30-day cost, conversions, CPA and search impression share lost to budget. Tell me which campaigns are budget-capped while beating my target CPA of $[X], and which spend more than 20% of the account above target.

A change, with approval (runRawMutate)

Prompt

Pause the ad groups that spent more than $300 in the last 30 days with zero conversions. Show me the list first and wait for my approval before making any change.

To make these repeatable, save them as Cursor skills for Google Ads — 15 ready-made SKILL.md files with the same prompts structured so they produce the same table every week, invoked with a slash command.

What to know before a heavy session

Every route inherits Google Ads API behavior, and Cursor adds a few constraints of its own. None of them block normal use, but they explain most confusing moments in the first week.

Project vs user config. ~/.cursor/mcp.json is not copied to Cloud Agents or remote workers. Put the entry in the project's .cursor/mcp.json when a team or a scheduled agent needs it — the managed connector holds no secrets in the file, so it is safe to commit.
Approval fatigue is a setting, not a fact. Cursor asks before every MCP call by default. Allowlist the read tools you use daily in Auto-review mode and leave writes gated.
Keyword Planner is rate-limited. Google allows roughly one planning request per second per customer ID, so a 500-seed keyword pull runs sequentially. Ask for batches of 20–50 seeds.
Write scopes are a separate decision. On Google's own server you choose scopes at OAuth time; on the managed connector every write tool is approval-gated regardless of scope. Decide before the first session which you want.
The Agent will happily write a script instead. If a tool call fails, Cursor sometimes offers to reach the API directly with Python. That needs a developer token you may not have — say no and fix the connector.

That last point is the real boundary of the analyst setup: Cursor over MCP tells you what to change and files the evidence, and you change it. If you want the changes shipped without you in the loop, Ryze AI's Paid Ads Autopilot runs bids, budgets and negatives 24/7 on the same account, at $89/month flat.

How we tested and scored the routes

How we tested — and who is writing this

What we tested: each route set up from a clean Cursor install against a live Google Ads account spending about $40K/month, September 1–3, 2026.
What we measured: time from opening mcp.json to a correct answer about live data, which read and write operations were available, what safety controls existed, and the cost to run it monthly.
Pricing: verified against each vendor's public pricing page on September 3, 2026. Ryze AI's connector is free to connect; its Paid Ads Autopilot is $89/month flat.
Disclosure: Ryze AI publishes this guide and operates the connector ranked first. The comparison data is shown so you can judge every route independently.

Each route was rated 0–5 on four criteria: coverage of the Google Ads API (×3), setup speed for a non-developer (×3), safety controls such as approvals and undo (×2), and cost (×2), for a component total out of 50.

Route	Coverage ×3	Setup ×3	Safety ×2	Cost ×2	Total / 50
Ryze AI managed MCP	5	5	5	5	50
Google's official MCP	5	2	3	5	37
Windsor.ai MCP	3	4	4	4	37
Agent-written API script	5	1	2	5	32
Google Ads Scripts via Cursor	3	2	3	5	31

The component totals are not the ranking by themselves. The editorial score also weights whether the route lets the Agent act on the account with a human approval in the loop, which is why the two read-only routes sit below Google's server despite tying it on components.

Alex M.

VP Growth
Multi-Brand Ecommerce Group

★★★★★

“

We had five channels and five people each defending their own budget. Handing allocation to Ryze AI cut blended CAC 27% in a quarter — not because the bidding got smarter, but because money finally moved to whichever channel was winning that week.”

Start free trial

27%

Blended CAC reduction

Channels automated

12 weeks

To full autonomy

1,000+ marketers use Ryze

Automating hundreds of agencies

★★★★★4.9/5

Get started

Which route should you pick?

The decision comes down to two questions: do you already hold a Google Ads developer token, and do you want the Agent to be able to change the account or only read it and write files.

Marketer, no developer token

Ryze AI managed MCP

Three lines in mcp.json, Google sign-in, live in about 2 minutes, write tools behind approvals. Safe to commit to the repo.

Developer with API access

Google's official MCP server

Runs as a stdio server on your machine; you choose scopes. Budget an hour for the Cloud project and OAuth client, and keep secrets in env vars.

Reporting only, many platforms

Windsor.ai MCP

Read-only marketing data across ad platforms in one endpoint. No writes, so no risk, but no changes either.

Custom pipeline or scheduled job

Agent-written API script or Google Ads Scripts

Let Cursor write the code, then run it where it belongs — a cron job for Python, the Scripts editor inside Google Ads for JavaScript.

Five-minute decision

Do you want the Agent to change anything? If yes, you need write tools — routes 1, 2 or 4.
Do you hold a developer token today? If no, route 1 is the only one under 10 minutes.
Will a Cloud Agent or a teammate run it? Then the config lives in the project's .cursor/ folder, not your home directory.
Do you also run Meta, TikTok or GA4? One URL that covers all of them beats one server per platform.
Do you want the changes executed without you? Then the connector is the diagnostic layer and an autopilot is the execution layer.

Whichever route you choose, pair it with a structured instruction set — the difference between a vague prompt and a saved skill is bigger than the difference between any two connection methods. Start with the 15 Cursor skills for Google Ads, and for the wider integration picture see the best MCP servers for Google Ads.

Frequently asked questions

Can Cursor connect to Google Ads?

Yes. Cursor's Agent supports MCP servers configured in .cursor/mcp.json or ~/.cursor/mcp.json. Add a Google Ads MCP server — the managed Ryze AI connector needs only a url and a Google sign-in; Google's own server runs locally but needs a developer token and a Cloud project — and the Agent can read and, with approval, change the account.

What is the fastest way to connect Cursor to Google Ads?

Put {"mcpServers":{"ryze":{"url":"https://connector.get-ryze.ai/mcp"}}} in .cursor/mcp.json, open Cursor Settings → Tools & MCP to confirm the server is on, and sign in with Google when the browser window opens. In our September 2026 run the Agent was answering about live campaign data in under two minutes.

Do I need a Google Ads API developer token to use Cursor with Google Ads?

Only on the self-hosted routes. Google's official MCP server and any script the Agent writes against the API require a developer token, a Cloud project and an OAuth client. A managed connector holds the API credentials on its side and authenticates you with plain OAuth 2.0, so no token application is needed.

Where does the Cursor mcp.json go?

Two places. .cursor/mcp.json in the project root is shared through Git and is the one Cloud Agents and teammates see. ~/.cursor/mcp.json in your home directory is personal and applies to every project on your machine but is not copied to Cloud Agents or remote workers. Use the project file for team workflows.

Can the Cursor Agent make changes to my Google Ads account?

It can, if the connector exposes write tools. The managed Ryze AI connector has six — apply or dismiss recommendations, raw mutate, delete, image upload and custom audiences — each gated behind your approval on the Ryze side, on top of Cursor's own per-call approval. Google's server executes whatever the OAuth scopes allow.

Is it safe to commit mcp.json to the repo?

For the managed connector, yes — the entry is a public URL with no secrets, and OAuth tokens live on the server side. For servers that need API keys or developer tokens, keep the values in environment variables using Cursor's env reference syntax and never commit them.

Does Cursor ask before running Google Ads tools?

Yes, by default Cursor asks for approval before every MCP tool call and shows the arguments. In Auto-review mode you can allowlist specific tools so reads run without a click. Keep write tools off the allowlist; on the managed connector they are approval-gated server-side as well.

Can Cursor run GAQL queries?

Yes, when the connector exposes a raw GAQL tool. Ask the Agent for campaign cost, conversions and impression share for the last 30 days and it writes and runs the query, then can save the result to a file in the repo. Remember that money fields come back in micros — divide by one million.

Can Cursor pull Keyword Planner data?

Through a server with the planning tools, yes. The managed connector exposes generateKeywordIdeas and generateKeywordHistoricalMetrics, returning monthly searches, competition and top-of-page bids. Google rate-limits planning calls to about one per second per customer ID, so request seeds in batches of 20–50.

Can Cursor Cloud Agents use the Google Ads connection?

Yes, if the server entry is in the project's .cursor/mcp.json and the server is remote. User-level configuration and local stdio servers are not available to Cloud Agents. The managed connector qualifies because it is a public URL; Google's local server does not without a custom worker image.

Cursor vs Claude for Google Ads — which connects better?

Both use the same MCP standard and the same server URL, so the connection is equivalent. Claude is the better chat analyst; Cursor's advantage is that the Agent writes files and code next to the data, so audits become CSVs, digests become Markdown and pipelines become scripts in the same repo.

What does connecting Cursor to Google Ads cost?

The connection itself is free on the managed route; Google's server is free to run locally but needs your credentials. MCP works on Cursor's standard plans. Ryze AI's Paid Ads Autopilot, if you want execution rather than analysis, is $89/month flat.

Related guides

15 Cursor Skills for Google Ads

SKILL.md files that turn the connection into a repeatable slash-command audit.

How to Connect Cursor to Meta Ads

The same mcp.json route for Meta — tools, steps and limits.

Best MCP Servers for Google Ads

Every Google Ads MCP server ranked, managed and self-hosted.

Top 10 AI Tools for Google Ads Management

When you want execution, not just analysis — the tools compared.

Ryze AI — Autonomous Marketing

Let AI run your Google Ads

✓Connect your account in minutes
✓The agent decides, applies and re-measures 24/7
✓Approval mode if you want the final say

Start free trial

2,000+

Marketers

$500M+

Ad spend

Countries

Ryze AI — Google Ads on autopilot

flat fee · free trial

Start free trial

How to Connect Cursor to Google Ads (MCP Setup Guide, 2026)

Free skills and prompts for paid ads and SEO

Quick comparison: the 5 ways to connect Cursor to Google Ads

What Cursor can and cannot do with Google Ads today

Step-by-step: connect Cursor to Google Ads with the managed MCP connector

1. Create the mcp.json entry

2. Confirm Cursor picked it up

3. Sign in with Google

4. Run the first prompt

5. Decide what the Agent may change

The 5 ways to connect Cursor to Google Ads, ranked

Ryze AI managed MCP

Google's official Google Ads MCP

Windsor.ai MCP

Stop pasting CSVs into the Agent — connect the account

Agent-written API script

Google Ads Scripts via Cursor

Get a free instant audit

Paid ads audit

SEO audit

The first 6 prompts to run once Cursor is connected

Account summary (getAccountSummary)

Wasted spend (runRawGaql on search terms)

Recommendations triage (listRecommendations)

Keyword research (generateKeywordIdeas)

Budget check (runRawGaql on campaign budgets)

A change, with approval (runRawMutate)

What to know before a heavy session

How we tested and scored the routes

Which route should you pick?

Marketer, no developer token

Developer with API access

Reporting only, many platforms

Custom pipeline or scheduled job

Five-minute decision

Frequently asked questions

Related guides

Let AI run your Google Ads

Let AIRun Your Ads

Let AI
Run Your Ads