Skip to main content
Model Context Protocol (MCP) standardizes the way large language models, AI apps, and agents connect to external systems. An MCP server exposes tools that an AI runtime can call when it needs fresh data or structured actions. For Sports MCP Server, that means your host does not need to learn a maze of low-level sports endpoints. It can call a hosted MCP server over HTTP, discover the tool catalog, and use sports-aware tools that already map to product jobs like scoreboards, match centers, season hubs, and watchability flows.

​
What You Need To Know First

TopicWhat it means here
TransportStreamable HTTP over POST /
Endpointhttps://sports-mcp-server.etonecarg.com/
AuthAuthorization: Bearer <API_KEY>
Health checkGET /
Core methodsinitialize, tools/list, tools/call

​
The Three MCP Calls You Will Use Most

MethodWhy it matters
initializestarts the session and declares client capabilities
tools/listreturns the tool catalog your host can call
tools/callruns a specific Sports MCP Server tool

​
Typical Host Flow

  1. Add the endpoint and bearer token to your MCP host or backend.
  2. Send initialize.
  3. Call tools/list.
  4. Start with a discovery, resolution, or bundle tool tied to your use case.

​
Example Host Configuration

{
  "transport": {
    "type": "streamable-http",
    "url": "https://sports-mcp-server.etonecarg.com/",
    "headers": {
      "Authorization": "Bearer YOUR_SPORTS_MCP_SERVER_API_KEY"
    }
  }
}

​
Example tools/call Wrapper

Sports MCP Server tool examples in the rest of the docs show the arguments payload. A direct HTTP request wraps those arguments like this: 
{
  "jsonrpc": "2.0",
  "id": "example-1",
  "method": "tools/call",
  "params": {
    "name": "gns_resolve_entity",
    "arguments": {
      "query": "Premier League",
      "type": "league",
      "limit": 5
    }
  }
}

​
Response Shape

Tool responses come back through MCP as text content containing JSON. That gives you two practical options:
  • let your MCP host unwrap and parse the payload for you
  • if you call the server directly over HTTP, read the text field and parse the JSON yourself

​
When To Use Atomic Tools Versus Bundle Tools

Choose thisWhen it is the better fit
Atomic toolsyou want precise control, smaller payloads, or multi-step agent reasoning
Bundle toolsyou want a product-ready response for dashboards, hubs, match centers, or historical views

​
The Most Common Evaluation Pattern

Most teams do not begin with a deep sports object ID already in hand. A strong evaluation usually looks like this:
  1. resolve a fuzzy user phrase with gns_resolve_entity
  2. fetch current schedule or standings with a discovery or league tool
  3. graduate to a bundle tool once the product shape is clear

​
Where To Go Next