Ask your assistant
about the football data.
Point Claude at the archive and ask in plain language. It calls the same endpoints you would, with your key, under your plan’s limits — and shows you what it called.
Connecting it
One URL and your API key. The server is remote, so there is nothing to install and nothing running on your machine.
https://footballsoccerapi.com/mcp
Settings → Connectors → Add custom connector, then paste the URL above. When it asks for authentication, use your API key as a bearer token.
If you would rather edit the config file directly, it lives at
~/Library/Application Support/Claude/claude_desktop_config.json on a
Mac and %APPDATA%\Claude\claude_desktop_config.json on Windows:
{
"mcpServers": {
"football": {
"url": "https://footballsoccerapi.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Restart Claude afterwards. The tools appear under the connector icon in the message box.
One command, from anywhere:
claude mcp add --transport http football \ https://footballsoccerapi.com/mcp \ --header "Authorization: Bearer YOUR_API_KEY"
Then ask it something. /mcp inside Claude Code lists what
connected.
MCP is an open protocol, so anything that speaks it can connect: Cline, Continue, Zed, Codex, and a growing list of others. What they have in common is the two things above — the endpoint URL, and your key as a bearer token.
The details differ per client and they change often, so rather than print steps here that go stale, check that client’s own documentation for “remote MCP server” or “HTTP transport” and give it:
url: https://footballsoccerapi.com/mcp header: Authorization: Bearer YOUR_API_KEY
Some clients want X-API-Key instead of a bearer
token. Both work — being fussy about which reads as “it does not
work”.
What it can reach
A curated set of tools rather than the whole API surface. Models choose badly from a long list, so the tools are named as verbs and the ones that need a cursor are left out — an assistant handed a cursor invents values for it.
Calls go through the public API with your key, so your plan, your rate limit and your quota all apply exactly as they do in code. Nothing here reaches the database directly.
Worth knowing
It uses your quota
An assistant can fire several tool calls to answer one question — more than a person would. If you hit the pace limit it is a limit, not an outage, and asking one broad question beats several narrow ones.
A free key reaches less
The same tools, on yesterday, today and the fixtures ahead, and without market prices. Enough to see whether this is useful to you; not enough to check a model against the market.
It will get things wrong
The data is ours and the reasoning is the assistant’s. It shows which tools it called and with what, so you can check — and it is worth doing on anything that matters.
Try it on a question you already know the answer to
That is the fastest way to find out whether it is any good. Ask it something you could check yourself, and see whether the tool calls make sense.