Fan-Out MCP Server

How to use

You use the Fan-Out MCP server by configuring it in your MCP client setup, then invoking analysis tasks to discover which queries your content answers and where gaps exist. The server supports content-only analysis, keyword-targeted hybrid analysis, and keyword-only analysis. When you run an analysis, you receive an interactive dashboard-style output that highlights coverage scores, per-query evidence, and prioritized recommendations to improve your content for Generative Engine Optimization.

In practice, you will generally choose between: - Content-Only analysis to understand what your article inherently covers - Hybrid analysis to target a specific keyword and generate 15–25 related query variants across several types - Keyword-Only analysis to focus purely on keyword variants for speed Each run returns a structured report with coverage details, evidence quotes, and concrete steps to close gaps.

How to install

Prerequisites you need before you begin are: - Claude Desktop (or any MCP-compatible client) - Anthropic API key - Node.js 18 or newer (for local build and tooling)

Follow these concrete steps to install and start using the Fan-Out MCP server with Claude Desktop. The fastest path is to configure Claude Desktop to load the Fan-Out MCP tool directly.

{
  "mcpServers": {
    "fanout": {
      "command": "npx",
      "args": ["-y", "@houtini/fanout-mcp@latest"],
      "env": {
        "ANTHROPIC_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Additional setup notes

Configuring the Fan-Out MCP server in Claude Desktop requires editing the Claude configuration to register the Fan-Out MCP tool as an MCP server. The example above demonstrates using npx to install and run the latest Fan-Out MCP package and to pass your Anthropic API key securely through an environment variable. After saving the configuration, restart Claude Desktop to apply changes. You will then be able to run the Fan-Out MCP tool from Claude Desktop’s tool menu and initiate analyses.

Troubleshooting

If you encounter issues starting or using the Fan-Out MCP server, try these steps:
- Confirm Claude Desktop is restarted after changes to the MCP configuration
- Verify the configuration file syntax has no trailing commas or invalid JSON
- Ensure the ANTHROPIC_API_KEY environment variable is set to a valid key
- Check that the Node.js version is 18+ and that npm or npx commands are available in your shell
- If performance seems slow, consider reducing the number of variant types or running in keyword-only mode to speed up analysis.

Tools and usage notes

The Fan-Out MCP exposes an analysis tool focused on content-gap analysis. You can run analyses in different modes: - Content-Only Analysis for general coverage assessment - Hybrid Analysis for keyword-targeted query variants across five primary types: Equivalent, Specification, Follow-Up, Comparison, and Clarification (with optional expansion to more types) - Keyword-Only Analysis to generate keyword variants without content inference for faster results Each query is evaluated with evidence quotes from your content, and results include coverage status, gaps, and recommended actions.

Example workflows

1) Quick content gap check using Content-Only Analysis to understand baseline coverage. 2) Hybrid Analysis for a targeted keyword, producing 15–25 variants and identifying gaps tied to that keyword. 3) Keyword-Only Analysis to rapidly enumerate variants when you’re focusing strictly on keyword space.

Notes on scope and limitations

This MCP emphasizes generating useful query variants and assessing coverage with evidence from your content. It is designed to work best with focused, topic-driven content such as technical articles or product pages. Longer analyses require more time, and more variant types increase the total number of queries and processing time.

Supported values and environment

Environment variables you may see in the configuration include ANTHROPIC_API_KEY for authenticating to Claude services. Ensure you keep this key secure and do not commit it to public repositories.

What you get back

Across runs you will obtain a coverage score, per-query evidence quotes, gap notes, and prioritized recommendations. You will also receive a detailed report that you can export for further analysis and sharing.

analyze_content_gap

Primary MCP tool that analyzes a URL to determine content coverage, generates query variants, and outputs a detailed report with evidence for each query.