DataForSEO MCP + Claude: Build Your Own AI SEO Agent

Claude can suggest keywords all day, but it doesn’t have fresh search data on its own. It doesn’t know your real rankings, search volumes or who you’re competing with in the results. In this guide I’ll show you how to connect the DataForSEO MCP server to Claude so it can pull that data itself, and then use it for three jobs: an SEO audit, a competitor and keyword-gap report, and an AI search visibility check. Setup takes a few minutes.

Disclosure: DataForSEO sponsored the video this article is based on.

Prefer to watch? See the full walkthrough on YouTube. I build the whole thing on screen.

What DataForSEO and its MCP server do

DataForSEO sells SEO data through APIs: keyword research, SERP results, rank tracking, backlinks, on-page analysis, competitor data and AI optimization data. Normally you’d write code to call those APIs. The DataForSEO MCP server wraps them as tools an AI assistant can call directly through the Model Context Protocol.

So instead of building API requests, you tell Claude what you want, such as “analyze this domain and find the biggest SEO opportunities”, and Claude decides which DataForSEO tools to call and fetches the data in seconds. The important part is that its recommendations now sit on real data you can check, not on guesses.

How pricing works

The MCP server itself is free and open source. You pay pay-as-you-go rates for the API calls Claude makes, with no monthly subscription. At the time of writing, DataForSEO’s site lists a $1 free trial credit, a $50 minimum top-up when you add funds, and credits that don’t expire. Example rates it lists include live SERP checks from $0.002 per query. Check the pricing page before you plan a big workflow, because deep requests cost more than basic ones.

If you want to know whether AI search mentions you at all before building the agent, read how to check if your site is visible to AI answer engines.

Step 1: Create a DataForSEO account

Sign up at DataForSEO and open the API Access tab. You’ll find your API login and password there. You need them for local setups. The hosted MCP server can use OAuth instead, so you approve access in the browser.

Step 2: Connect the MCP server to Claude

How you connect depends on which Claude you use. DataForSEO’s current docs give these options:

Claude desktop app

1. Open Customize → Connectors.
2. Click Add custom connector and paste the hosted MCP URL: https://mcp.dataforseo.com/v3/mcp.
3. Approve the authorization request.
4. Start a new chat. Claude asks for approval the first time it calls a DataForSEO tool.

Claude Code

You can add the server locally from npm with your API credentials:

claude mcp add dfs-mcp --env DATAFORSEO_USERNAME=<api_login> --env DATAFORSEO_PASSWORD=<api_password> -- npx -y dataforseo-mcp-server

DataForSEO’s help center also documents a remote option and a custom local build. Whichever you choose, restart Claude afterwards so it loads the connection, then check that the DataForSEO tools appear.

Step 3: Run an SEO audit

Start with one site. I used my own, TubeAnalytics. A prompt like this works:

Use DataForSEO to analyze example.com. Run an SEO audit and identify the five most important SEO issues or opportunities to investigate first.

You’ll see Claude make tool calls, then return its findings. Don’t just read the recommendations. Look at the evidence underneath, such as pages with technical issues, ranking keywords and SERP data, because that’s what separates this from asking a model to guess.

Then turn it into a plan:

Prioritize these opportunities by potential business impact and difficulty, then give me a 7-day action plan.

If you want help turning outputs like this into repeatable workflows, the Claude Code courses in AI Income Lab cover building with skills and MCP servers step by step.

Step 4: Build a competitor and keyword-gap report

Use DataForSEO to create a competitor SEO report for example.com. Identify relevant search competitors, important overlapping long-tail keywords and useful keyword gaps. Summarize the findings in a table.

Claude returns competitors, overlapping keywords and gaps. Follow up with something like “From these gaps, find five realistic content opportunities and explain why each one is worth investigating.” Because you’re in a conversation, you can keep asking questions, and Claude goes back to the data each time.

Step 5: Check AI search visibility

More businesses now want to know whether ChatGPT, Claude, Gemini and Perplexity mention and recommend them, not only where they rank in Google. DataForSEO has AI optimization APIs for that kind of research, and Claude can call them through the same connection.

Use the available DataForSEO tools to investigate AI search visibility for example.com. Identify signals that show how visible this brand is in AI-driven search, and summarize what I should investigate next.

What comes back depends on how much data exists for your domain and which APIs Claude chooses. Treat it as another research layer alongside your normal SEO work, not a replacement.

Start small

Don’t build a huge autonomous SEO agent on day one. Connect the MCP server, pick one website, and test three prompts: an SEO audit, a keyword-gap report and an AI visibility check. Watch which tools Claude calls and what comes back. Once you trust the output, turn the prompts that work into a repeatable workflow or a Claude skill.

For other ways to automate this, see this Claude skill for SEO work and how I built an AI search visibility tracker.

Common mistakes to avoid

• Not restarting Claude. New MCP connections often don’t appear until you restart or start a new chat.
• Skipping the evidence. Check the data behind each recommendation before acting on it.
• Running open-ended prompts on big sites. Deep requests cost more. Start narrow and watch your balance.
• Mixing up credentials. Use the API login and password from the API Access tab, not your website login.
• Trusting AI visibility data blindly. It’s a useful signal, not a ranking guarantee.

If you get stuck connecting MCP servers, you can ask in the AI Income Lab community while you build.

FAQ

Conclusion

Connecting DataForSEO to Claude turns it from an idea generator into an assistant that works with real search data. Set up the MCP server, run the three starter prompts on one site, check the evidence, and build a workflow from the prompts that earn their place. Watch the video above to see each step on screen.