KeywordGraph MCP Server

What it is

The KeywordGraph MCP server is a Model Context Protocol endpoint that connects large language models — Claude, ChatGPT, Cursor, and any MCP-capable client — directly to KeywordGraph’s knowledge-graph engine. Instead of guessing at keywords, the model pulls live search data, builds topic graphs, and runs gap analysis through the same algorithms behind the web app.

This page covers the SEO and LLMO side of the server: the tools that matter when the job is ranking on Google or getting cited by AI search engines.

Where it connects

• Claude (web, desktop, and Claude Code)
• ChatGPT (Deep Research and custom connectors)
• Cursor and other MCP-aware IDEs
• n8n and Make.com automation workflows
• Local LLM clients and the terminal

SEO & LLMO tools

The tools below are the ones tuned for search and generative-engine optimization. Each returns structured graph data the model can reason over, not just a flat list.

• analyze_google_search_results — pull the live SERP for a query and turn it into a topic graph, so you can see what the ranking pages collectively cover.
• analyze_related_search_queries — map the search-demand side: the related questions and queries people actually run, clustered by intent.
• search_queries_vs_search_results — overlay demand against supply to surface the content gaps: topics people search for that the current results under-serve.
• generate_content_gaps — find the structural holes in any text or topic and get the bridges that close them — the core LLMO move for becoming the source an AI engine cites.
• generate_topical_clusters — cluster keywords and entities by topical community to plan pillar-and-spoke coverage.
• generate_seo_report — a full search-intent and content-gap report for a topic, built from real Google data in one call.
• create_knowledge_graph — build an entity graph from a document or URL to audit topical coverage and internal-link structure.

How to connect

Two ways in, depending on the client.

• Remote (recommended). Point your client at https://mcp.keywordgraph.com and authorize with OAuth. Nothing to install.
• Local. Run the server locally and authenticate with a KeywordGraph API key from your account settings — useful for terminal and self-hosted setups.

To set up with a local LLM client using a locally stored API key, add the KeywordGraph MCP server to your client config and replace your-api-key-here with your own key:

{
  "mcpServers": {
    "keywordgraph": {
      "command": "npx",
      "args": ["-y", "keywordgraph-mcp-server"],
      "env": {
        "KEYWORDGRAPH_API_KEY": "your-api-key-here"
      }
    }
  }
}

Either way the model gains the same SEO and LLMO toolset, and any data it pulls — SERPs, search intent, gap analysis — augments its answers with structure rather than guesswork.

Connect the MCP server

A KeywordGraph account gives you the remote MCP endpoint plus the full web app. Start free and wire it into Claude or ChatGPT in a couple of minutes.