Give Claude live tennis data with MCP

Connect Claude, Cursor or any MCP client to real-time tennis scores, odds and model win-probability with one command — 12 read-only tools over the Live Tennis API.

Large language models do not know today's tennis scores. The Model Context Protocol fixes that by giving the model tools it can call. This connects Claude to live tennis data in one command.

Setup

Claude Code

claude mcp add livetennis -e LIVETENNISAPI_KEY=twjp_… -- npx -y livetennisapi-mcp

Claude Desktop — add to claude_desktop_config.json:

{
  "mcpServers": {
    "livetennis": {
      "command": "npx",
      "args": ["-y", "livetennisapi-mcp"],
      "env": { "LIVETENNISAPI_KEY": "twjp_…" }
    }
  }
}

Cursor, Zed and any other MCP client take the same command and the same environment variable. There is nothing to install — npx fetches it on demand.

Then just ask

What tennis matches are live right now?

Who's winning the Alcaraz match, and what does the model give him?

Show me Sinner's ranking and recent results.

The 12 tools

Tool	Returns	Plan
`get_live_matches`	Matches in progress with live scores	BASIC
`get_upcoming_matches`	Matches starting soon	BASIC
`get_match`	Full detail for one match	BASIC
`get_match_score`	Current score only — fastest read	BASIC
`search_players`	Find players by name	BASIC
`get_player`	Profile, ranking, country, handedness	BASIC
`get_fixtures`	Forward schedule	BASIC
`get_recent_results`	Completed matches and winners	BASIC
`get_match_events`	Breaks, games, sets, momentum runs	PRO
`get_match_odds`	Match-winner prices, bid / ask / mid	PRO
`get_match_analysis`	Model thesis and win probability	ULTRA
`check_api_status`	Reachability and which plan your key is on	—

Everything is read-only. No tool can modify anything.

A design decision worth explaining

The API gates endpoints by plan and returns a bare 403 {"error":"upgrade_required"}. Hand that to a language model and it will usually invent a reason or retry pointlessly — neither of which helps you.

So every tool that can hit a plan wall returns a plain-English explanation instead, as a normal result rather than an error:

This data requires the ULTRA plan, and the configured API key is on a lower tier. Nothing is wrong with the key — the endpoint is simply not included in the current plan.

Marking it as an error makes MCP clients retry or bail, which loses the one piece of information you actually need. check_api_status also probes upward to report which plan your key is really on, so you can diagnose it without guessing.

Your key stays local

The server runs on your machine. The key is read from the environment by that process and is sent only to api.livetennisapi.com.

Where to go next

Source: github.com/livetennisapi/livetennisapi-mcp
npm: livetennisapi-mcp
Official MCP Registry: io.github.livetennisapi/livetennisapi-mcp
Plans: https://livetennisapi.com/#pricing