Mike HolpJoin AI Income Lab ↗

Build guide · By · Published

Give Claude Code real SEO data with DataForSEO.

Ask an AI model which keywords your site should target and you get ideas, not data. The DataForSEO MCP server lets Claude Code call DataForSEO’s APIs for search results, keyword metrics and site analysis, so its answers rest on numbers you can check. This guide connects it and runs one small query you can verify. It pairs with my DataForSEO and Claude video.

DataForSEO sponsored that video. This guide uses only their public docs and plain links, with no referral codes. You need Claude Code and a DataForSEO account. DataForSEO charges per API request rather than a monthly plan, so each query costs money once any trial credit is used; check their current pricing before you start.

Connect the MCP server, confirm Claude Code lists its tools, and run one small query on a site you own.

1. Get your API credentials

Create a DataForSEO account, then open API Access in the dashboard. It shows an API login and an API password. These are not your website login: they are what the MCP server uses to make paid requests, so treat the password like a card number.

2. Connect the MCP server to Claude Code

The official DataForSEO MCP server runs locally through npm and needs Node.js 22 or newer (node -v to check). Add it with your credentials as environment variables, replacing the two placeholders:

claude mcp add dataforseo \
  -e DATAFORSEO_LOGIN=your_api_login \
  -e DATAFORSEO_PASSWORD=your_api_password \
  -- npx -y dataforseo-mcp-server

Claude Code stores these values in its own configuration on your computer, so don’t run this on a shared machine or paste the command anywhere public. DataForSEO also documents a hosted server at https://mcp.dataforseo.com/v3/mcp that signs in with OAuth instead of a stored password. Their Claude Code help page lists the options.

3. Confirm the tools load

From a terminal:

claude mcp list

dataforseo should appear as connected. Then start Claude Code and type /mcp: the DataForSEO entry should list its tools, such as docs_index for browsing the API documentation. If it shows as failed, check node -v and the credentials before anything else.

4. Run one small, checkable query

Start small so you can see what each call costs and returns. Use a site you own, and ask Claude to name the tools before it calls them:

Use DataForSEO to find the ten keywords example.com ranks for in the United States,
with position and search volume. Tell me which DataForSEO endpoint you will call
before calling it, and keep it to one request.

Replace example.com with your domain. Claude Code will ask permission before the first DataForSEO tool call; read which endpoint it chose before you approve. Then check the result two ways:

If those match, the connection is real and you know what a query costs. That is the point to stop and decide whether it’s worth more.

5. Grow it into a workflow

In the video I ran three kinds of prompt against my own site, TubeAnalytics. Each one is useful once the small query works:

Larger prompts can trigger many API calls. Ask Claude to estimate how many requests a prompt needs before it runs, and keep prompts that work in a file so you can repeat them on a schedule instead of exploring from scratch each time.

When it fails

Setup checked October 22, 2026 against dataforseo-mcp-server 3.1.3, the server’s README, DataForSEO’s Claude Code help page (updated October 16, 2026) and Claude Code 2.1.307. Prices, trial credits and endpoints can change; check DataForSEO before relying on them.