MCP (Model Context Protocol)

Weight: 10% of your AX score. The Model Context Protocol (MCP) configuration at /.well-known/mcp.json describes tools and resources that AI agents can interact with programmatically.

mcp.json not found

Create a /.well-known/mcp.json file describing your MCP server:

{
  "name": "Your Site MCP Server",
  "description": "Provides tools for querying site data and resources.",
  "protocolVersion": "2024-11-05",
  "tools": [
    {
      "name": "search",
      "description": "Search site content by keyword.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "query": { "type": "string", "description": "Search query" }
        },
        "required": ["query"]
      }
    }
  ],
  "resources": [
    {
      "name": "site-data",
      "description": "Structured site data in JSON format.",
      "uri": "https://your-site.com/api/data"
    }
  ]
}

Not sure if you need MCP?

MCP is most valuable if your site has an API or interactive features. If your site is purely static content, the other AX files (llms.txt, agent.json) are more important. You can still create a minimal mcp.json to improve your score.

Invalid JSON

Your mcp.json has invalid JSON syntax. Validate it with a JSON linter.

Missing server name

Add a "name" field identifying your MCP server. This helps AI agents distinguish your server from others.

Missing server description

Add a "description" field explaining what your MCP server does and what tools it provides. This is the primary text AI agents use to decide whether to interact with your server.

Tools missing descriptions

Some of your tools don't have a description field. Add descriptions to every tool so AI agents understand what each one does and when to use it.

No tool descriptions

None of your tools have descriptions. Without descriptions, AI agents can't determine which tool to call for a given task. Add a clear "description" to each tool entry.

Empty tools array

Your tools array exists but is empty. Add at least one tool with name, description, and inputSchema fields. See the template above for an example.

No tools array

Your mcp.json has no tools array. This is the most important part of the MCP configuration — it tells AI agents what operations are available.

No resources defined

Add a "resources" array listing data sources your MCP server exposes. Resources are read-only data endpoints that AI agents can access for context.

No protocol version

Add "protocolVersion": "2024-11-05" to declare MCP spec compatibility. This tells AI agents which version of the protocol to use when communicating with your server.

No CORS on MCP endpoint

Browser-based AI agents need CORS headers to fetch your mcp.json cross-origin. Add Access-Control-Allow-Origin: * to the response. See the HTTP Headers guide for implementation details.