A local car detailing business owner in central Texas had Claude Code research, write, and set up an entire Google Ads search campaign, and the data costs came to $0.46. Here is what the agent actually did, what each step cost, and the guardrails worth copying. (h/t the owner who shared the original write-up.)
The point is not that ads are now free. The point is that the research layer that used to need an expensive subscription can now cost less than a dollar per campaign.
The problem: ads that didn't bring clients
The owner used to build ads by hand: look up keywords in Google's keyword tool, check what competitors were running, write the campaigns. According to the author, those campaigns weren't bringing in any clients. They then tried Semrush's cheapest plan, found it too limited, and quit.
First try: Claude Code without data
With only a Claude Code Pro subscription and no ad tools, the agent produced a generic plan from what it already knew:
- Mobile detailing: mobile car detailing, mobile auto detailing killeen, mobile detailing near me
- Interior: interior car detailing, car carpet cleaning, pet hair removal car
- Truck/SUV: truck detailing killeen, suv detailing, lifted truck wash
- Engine/tires: engine bay cleaning, engine detailing
- Premium: paint correction killeen, ceramic coating killeen, car scratch removal
- City ads: the same set for nearby towns
No search volumes, no costs, no competitor data. It looked like a plan, but nothing in it was verified.
Second try: Claude Code with an MCP and real data
The owner then connected an MCP server (the AIML API MCP) that promised an optimized campaign for under $1. The prompt was simple: connect the MCP and create an optimized ad campaign for the company, with the website link pasted in. Claude Code already had access to the Google Ads account through the browser.
When asked what it had done, the agent listed every step and its cost:
| Step | Tool | Calls | Cost |
|---|---|---|---|
| Check the landing page | DataForSEO on-page parsing and Lighthouse | 1 each | $0.007 |
| Keyword ideas from the site | DataForSEO keywords for site | 1 | $0.117 |
| Keyword ideas from 20 seed terms | DataForSEO keywords for keywords | 1 | $0.117 |
| Find rival advertisers | DataForSEO SERP advertisers | 6 | $0.009 |
| Pull their ads | DataForSEO SERP ads search | 6 | $0.009 |
| Check which ads show on live searches | Litescrape Google Ads | 10 | $0.002 |
| Sort keywords into buyers, other brands, junk | Typesafe router | 2-4 | ~$0.05 |
| Forecast clicks and cost | DataForSEO ad traffic | 1 | $0.117 |
| Write ad groups, ads, negative keywords | Typesafe router | 1 | ~$0.03 |
It then set up the whole campaign in Google Ads and asked for the owner's OK before launching.
The keyword plan it produced
The most useful output was a tiered keyword list with Texas search volume and CPC.
Tier 1 (core terms, most of the budget):
| Keyword | TX searches/mo | CPC |
|---|---|---|
| car detailing near me | 27,100 | $3.03 |
| auto detailing near me | 6,600 | $2.59 |
| mobile detailing near me | 5,400 | $3.98 |
| mobile car detailing / mobile auto detailing | 2,900 each | $5.48 |
| interior car detailing | 1,300 | $3.78 |
| interior detailing near me | 590 | $4.05 |
| car detailing killeen | 320 | $2.33 |
| car detailing temple tx | 170 | $1.86 |
A sharp caveat from the agent: many people searching "car detailing near me" want a shop, not a mobile service. It recommended putting "mobile" and "we come to you" in the ad copy and watching the search terms report closely.
Tier 2 (cheaper and useful): truck detailing (210 searches, $3.05), engine bay cleaning (720, $1.30), car wash and detail (2,400, $1.96), mobile car wash (1,600, $2.68), and price-shopper terms like car detailing prices and mobile detailing cost (720 each, $1.58). Price-shopper terms only work once the website shows prices.
Tier 3 (test small, only if you offer it): paint correction near me (1,000, $3.53), paint correction (1,300, $4.42), ceramic coating near me (2,900, $8.28).
What the community added
One commenter pointed out local context the data could miss: Salado has around 2,000 people, so few residents are likely searching for ceramic coating there, while Killeen's market is shaped by the nearby military base. Keyword data tells you what people search, not who your best customers are.
Honest limits
The owner was clear: the $0.46 covered putting the campaign together. Whether it brings in clients is the next thing to measure.
- Ad spend is separate. The $0.46 is research and setup cost, not your click budget.
- You still need a Claude Code subscription and access to the data tools through an MCP.
- Giving an agent your ad account is a real risk. Keep a human approval step before launch.
- Forecasts are not results. Volumes and CPCs are estimates.
How to do it
- Connect your accounts. Give Claude Code access to your website and Google Ads account.
- Connect a data MCP. The owner used an MCP that routes to DataForSEO, Litescrape, and a model router for sorting and writing.
- Ask for tool prices first. Have the agent list each tool it plans to call and the cost before running anything.
- Run the campaign prompt. Paste your website link and ask for an optimized ad campaign.
- Demand the data behind every suggestion. Search volumes, CPCs, competitor ads, and why each keyword made the cut.
- Edit with local knowledge. Remove towns, services, or price terms that don't fit your real market.
- Review before launch. Check keywords, ads, negative keywords, and budget before approving anything.
- Measure clients, not clicks. Track calls, bookings, and the search terms report, then adjust.
Originally published on SoloAgentFlow.
