An SEO tool for $50 instead of €480 a month. The real cost.

A CUVIR client wanted to know what his three biggest competitors are doing online. Not as a gut feeling. As data. Keyword by keyword: who ranks where and how much traffic lands there in real numbers.

The standard route: a Semrush subscription. The Business plan. I saw the price on the screen and paused for a moment.

€479.45. Per month.

For a single client and a single question, that would be a subscription I need for two weeks and keep paying for eleven months. So I took a different route. I'm building the thing myself.

Why the expensive subscription isn't the right answer for many people

Semrush, Sistrix, Ahrefs are good tools. For an agency with thirty active SEO clients, the subscription pays for itself in an afternoon. That's not the question.

The question is: what about someone who only needs SEO now and then? The freelancer who rebuilds a website for three clients a year. The operator who wants to know whether his own product shows up on Google and in AI answers at all. For him, a full subscription is a size too big.

And there's a second point that interests me more than the price. Semrush is a closed interface. I can look inside. I can't pass the data on to my own system without a detour: to an agent, to a script that turns the data straight into an action. That's where the lever is.

The mechanics: data and interface are two different things

The mistake in thinking I made myself at the start: I thought an SEO tool is one thing. It isn't. It's two.

The first is the data. Who ranks for which keyword, how high the search volume is, what a competitor's site structure looks like. Nobody collects this data by hand. It comes from a crawler service that goes through the web and provides the numbers. One of these is called DataForSEO. It doesn't sell an interface. It sells raw data through an API.

The second is the interface. What you see when you type "Immobilien München" and get a sorted list back. That's exactly what's missing when all you have is the raw API. I tried it: I connect the DataForSEO API directly to an AI coding tool, ask a question and get a wall of JSON back. Technically correct. Useless as a tool for a client call.

Now the free repo OpenSEO enters the picture. It's the missing interface. It takes the data from DataForSEO and turns it into a dashboard that looks structurally like the expensive tools. Keyword research, domain overview, search intent, site audit. The interface costs nothing because it's an open project. You only pay for the data you pull.

That's the whole mechanism. A free interface plus usage-based data plus a connection through which my own agent accesses both.

The process, the way I ran it

I didn't configure this by hand. I had it set up for me. If you have an AI coding tool, give it the following process and let it do the work.

1. Clone the repo: Tell the coding tool to fetch the OpenSEO repository from GitHub
 and set it up locally. The setup guide is in the repo, the tool reads
 it on its own and works through it.
2. Create a DataForSEO account: Sign up, add a small balance (I tested
 with $50), request API credentials. They arrive by email.
3. Hand over the credentials: Paste the DataForSEO login details into the setup.
 From here on, OpenSEO pulls its data from there.
4. First test: Type in a keyword. IMPORTANT: In keyword mode, choose "Suggest",
 not "Auto". The quality of the results is noticeably better with "Suggest".
5. Cross-check: Compare the results with a real Google search and, if available,
 with a well-known tool. Don't trust them blindly.
6. For automation: Connect OpenSEO to your coding tool via MCP so that an
 agent can access the SEO data directly.

A practical note on choosing the model, which saved me quota: for the setup alone, a strong model in normal mode is enough. I deliberately didn't burn the most expensive reasoning setup on it, because a repo setup isn't deep thinking. It's diligent grunt work that the AI handles cleanly anyway.

An example, calculated from start to finish

Back to the client who wanted to see his competitors.

I go to the domain overview and enter a large reference domain from his industry. In the video it was Engel & Völkers. I get organic traffic, the number of ranking keywords, the distribution of the rankings. A picture in thirty seconds, which would otherwise have needed a full subscription.

Then I turn it around. Instead of only looking, I use the data as a door opener. I take a potential client's competitor data and write to him, not with a sales promise, but with an observation. "Your competitor ranks for these twelve transactional search terms, you rank for none of them." That's not a pitch. That's a mirror.

And here the search intent feature kicks in, and it makes all the difference. The tool sorts keywords by intent. Informational means: someone wants to know something. Transactional means: someone wants to buy something. Pukaj offers a sample calculation that I'll leave standing as a rough rule of thumb, not as proof: a hundred clicks on transactional keywords can bring four or five sales. A thousand clicks on purely informational keywords can bring nothing at all.

That's exactly the argument a client understands without me having to explain SEO to him. Not reach. Purchase intent.

Where the savings pitch crumbles

Now the part that the original rushes through.

DataForSEO isn't free. It's usage-based. The claim that $50 lasts several months holds for my test case: a few clients, occasional queries. The community even says "easily half a year". For this usage profile, I believe it.

For an agency with many clients and daily bulk queries, the math is different. Every domain analysis, every site audit, every keyword query costs money. It scales with usage, and the video brushes this aside with a vague "should last a long time". If you run a lot of queries, at some point you end up with an amount that gets close to the €480 mark. Then the advantage is gone, and on top of that you're stuck with the maintenance of a self-hosted tool. Updates, API outages: you have to catch those yourself.

Second point: accuracy. Pukaj cites 80% accuracy for the traffic estimates. No source, no method. That's a ballpark figure, not a measurement. The numbers are based on mathematical crawler models, not on real analytics data from the other site. That's enough for a rough look at the competition. For a calculation that a budget depends on, I wouldn't rely on it.

Third point, and I consider it the most important: in the video, the comparison with the expensive tools runs on a single keyword. "Immobilien München", a German mass-market term that every tool has good data for. With niche keywords, with small language markets, with anything where the data gets thin, things can look completely different. One example is not proof.

And the claim that SEO is less contested today than ten years ago because everyone has moved to social media, I just don't believe. The SEO market is growing. What is true: the search surface has shifted. Besides Google, the answer engines now rank too, ChatGPT and the like. That's not an emptier market. That's a second market right next to it.

How I solved it in my own setup

For me, the value isn't in saving money. It's in the last line of the process above: the MCP connection.

A closed tool is a dead end. I look inside, I copy something down, I continue by hand. An open tool with an MCP connection is attached to my agent. The agent can pull the SEO data, derive a site structure from it and build me a proposal for a website relaunch right away. Data and action in one flow, with no copy-paste in between.

I'm building this blog article on the same principle. The AEO structure that answer engines prefer doesn't come from an SEO dashboard. It comes from my setup knowing how an article has to be built so that ChatGPT cites it. A tool that shows me a list doesn't replace that. A tool that's connected to my agent becomes part of it.

That's why I wouldn't switch for the number on the screen alone. For the open connection to my own system, I would.

FAQ

Does DataForSEO cost nothing at all?

No. DataForSEO is billed by usage, you pay per data request. Only the OpenSEO interface is free, a free GitHub repository. According to Pukaj, $50 in credit lasts several months for occasional use, and his community says "easily half a year". With heavy agency use and many daily queries, the costs rise accordingly.

What exactly is OpenSEO?

OpenSEO is a free GitHub repository that provides the missing interface for raw SEO data. It takes the data from a crawler service like DataForSEO and displays it as a dashboard, with keyword research, domain overview, search intent and site audit. The data itself doesn't come from OpenSEO. It comes from the connected API.

Why isn't it enough to connect the API directly to an AI tool?

Because the raw API doesn't give you a clear interface. If you connect DataForSEO directly to an AI coding tool, you get structured raw data as the answer: technically correct, but useless for a client call or a quick assessment. OpenSEO provides exactly the interface that's missing in between.

How accurate are the traffic numbers?

Pukaj cites about 80% accuracy, but without a source or methodology. The numbers are based on mathematical crawler models, not on the real analytics data of the analyzed site. That's enough for a rough look at the competition. For decisions that a budget depends on, the estimates are too uncertain.

What's the difference between informational and transactional keywords?

An informational keyword means someone wants to know something. A transactional keyword means someone wants to buy something. The difference decides what the traffic is worth: according to Pukaj's sample calculation, a hundred clicks on transactional keywords can bring four to five sales, while a thousand clicks on purely informational keywords can bring no results at all.

Build it yourself

• OpenSEO GitHub repository, a free repo, the interface for building your own SEO dashboard. URL announced in the video, but not known.
• DataForSEO, an API service for data on keywords, SERPs and domains. Create an account, add a small balance, request credentials.
• MCP (Model Context Protocol), an interface for connecting OpenSEO to an AI coding tool and feeding the data straight into automation.