Connecting football data to AI assistants with MCP

The connection exposes a small set of read-only tools, each mapped to a qualified feed. A client can ask for a competition's standings, a team's profile or a player's record by identifier, and receive structured data back. Tools never write anything, never place transactions and never reach feeds that are not enabled for your account.

Calls through MCP count towards your usage like any other request, and they appear on your usage page.

Our MCP documentation lists the current tools. If the server is unreachable, the documentation shows a fallback list so you can still plan.

From the connect AI page in your dashboard, you start an authorisation flow from your client. You sign in, review the requested access on a consent page and approve it. The client receives a scoped token tied to your account. You can see and revoke connections at any time from the same page.

Revocation takes effect immediately. If you stop using a client or suspect misuse, revoke and reconnect later if needed.

AI clients can summarise data fluently and also make mistakes. Encourage users to ask for the underlying figures, check identifiers and treat summaries as drafts. When a feed is not available, the tool returns a clear error, and a well-behaved client should say that the data is unavailable rather than inventing it.

The data contains no predictions of results, and the tools do not offer tips. A client asked who will win should be able to show tables and recent results, but any forecast is the client's own speculation, not FootballAPI.net data.

The connection only reads football data. It does not expose your API keys, billing details or other account information to the client. Tokens are scoped to the MCP tools and can be revoked. Review what your chosen client does with tool results, since its own privacy policy governs that.

AI clients can make several data calls per question. MCP requires Pro, which includes 250,000 successful requests per day; normal account usage rules apply. The inclusion resets at midnight UTC, with metered extras at €0.002 each and a shared safety limit of 100 requests per second. Cache reusable results and monitor usage after connecting a new client. A large daily allowance is not permission for unlimited concurrent calls.

The site's Ask AI box uses an AI model grounded in public guides, published endpoint parameters and current product settings. It can explain integration steps and cite documentation, but it cannot inspect your private invoices, recover keys or change an account. MCP is different: it lets a compatible external AI client call permitted read-only football tools after you authorise the connection.

Analysts can ask for a team's season statistics and get structured figures to explore. Editors can check a table or a fixture date quickly. Developers can prototype queries before writing code. In each case the value comes from fast, accurate lookups, not from the client's opinions.

An editor connects a compatible client, approves the consent page and asks for the current table of league 1204. The client calls the standings tool, receives the rows and formats them. The editor then asks for team 9260's profile, and the client calls the team tool. Both calls appear on the usage page. At the end of the week, the editor revokes the connection from the dashboard.

Specific questions produce better and cheaper results. Ask for the standings of a named competition and season rather than asking how a league is going. Ask for a team by name and let the client confirm the identifier. When the answer includes figures, ask the client to show the raw rows it used, which makes errors easy to spot.

If the client cannot find a competition, check the directory yourself and give it the identifier. Clients sometimes guess identifiers, and a wrong guess can return a valid response for the wrong competition.

Avoid open-ended requests to analyse everything about a club. They trigger many tool calls, consume usage quickly and rarely produce better insight than a few targeted lookups.

If several colleagues use AI clients, each should connect under their own account so usage and revocation are individual. Agree internal guidance on how AI summaries may be used in published work: for example, that every figure must be checked against the stored data before publication. The connection is a fast research tool, not a substitute for editorial checks.

Review active connections monthly and revoke any that are no longer needed.

A tool may return an error because a feed is not enabled for your account, a parameter is wrong or a limit was reached. Good clients relay that message plainly. If yours instead produces an answer anyway, treat the answer as unreliable and check the usage page for the failed call, which records the status code returned.