nutmeg

IN BETA!

Football data analytics for AI coding tools

Makes your AI agent an expert at football data. It looks up the real docs instead of guessing — qualifier IDs, coordinate systems, API endpoints, method signatures. All verified, all sourced.

Get started

Answer two questions and we'll show you exactly what to install.

What AI coding tool do you use?

What do you want?

Start here

/nutmeg

Describe what you want — it handles the rest

/nutmeg-learn

Concepts, resources, provider docs

Sub-skills auto-dispatched or called directly

/nutmeg-acquire

Fetch and download football data

/nutmeg-wrangle

Transform and reshape data

/nutmeg-compute

Derived metrics (xG, PPDA...)

/nutmeg-analyse

Explore and interpret data

/nutmeg-brainstorm

Visualisation ideation

/nutmeg-review

Check data code and charts

/nutmeg-store

Storage and publishing

/nutmeg-heal

Fix broken scrapers and pipelines

Providers

Provider	Events	Stats	xG	Free?
StatsBomb				Open data
Opta
Wyscout
SportMonks				Free tier
FBref
Understat

football-docs

IN BETA!

Like Context7 for football data

Searchable documentation across 21 football data providers and tools, built for AI coding agents. Covers club and provider APIs including StatsBomb, Wyscout, SportMonks, Impect, SkillCorner, Sportradar, TheSportsDB, BeSoccer, FMDB Pro, TransferRoom, Opta, and public/free sources, plus open-source libraries people build with too (kloppy, mplsoccer, socceraction, soccerdata, databallpy, floodlight, and more). Your agent resolves provider names, looks up the real docs, compares provider concepts, and resolves football entities across provider IDs instead of guessing from training data. Works with any MCP client.

1356

searchable chunks

providers + tools

MCP tools

v0.8.0

npm version

GitHub repo npm package

Install

Claude Code

Adds football-docs to the current Claude Code project.

claude mcp add football-docs -- npx -y football-docs@latest

Claude Code user scope

Use this when you want football-docs available across all Claude Code projects.

claude mcp add --scope user football-docs -- npx -y football-docs@latest

MCP JSON

Use this shape in clients that accept MCP server JSON config.

{
  "mcpServers": {
    "football-docs": {
      "command": "npx",
      "args": [
        "-y",
        "football-docs@latest"
      ]
    }
  }
}

Requires Node 20+. Restart your client, then ask use football-docs to look up Opta qualifier 76.

Tools

search_docs

Full-text search across all providers

resolve_provider_id

Resolve provider names and aliases

get_provider_docs

Retrieve docs for a resolved provider

list_providers

See all indexed providers

compare_providers

Compare how providers handle concepts

request_update

Flag outdated docs or request new ones

resolve_entity

Map players, teams, and coaches across provider IDs

Example queries

"What is Opta qualifier 76?"
"How does StatsBomb represent shots?"
"Compare Opta and Wyscout coordinates"
"Does SportMonks have xG?"
"How does SPADL represent a tackle?"

Indexed providers

StatsBomb

235 chunks

Wyscout

161 chunks

floodlight

145 chunks

kloppy

126 chunks

SportMonks

82 chunks

Impect

77 chunks

Opta

71 chunks

mplsoccer

64 chunks

databallpy

63 chunks

Free sources

62 chunks

SkillCorner

48 chunks

TransferRoom

43 chunks

soccerdata

40 chunks

FMDB Pro

35 chunks

socceraction

34 chunks

Sportradar

29 chunks

TheSportsDB

18 chunks

BeSoccer

14 chunks

FotMob

3 chunks

Soccerdonna

3 chunks

Transfermarkt

3 chunks

What's coming next release

v0.8.1 in development — not yet on npm

fast-forward and unravelsports in review

Both packages from UnravelSports, suggested by their author. fast-forward is a Rust-cored tracking loader covering 12 providers — six of which nothing in the index currently documents at all: Hawk-Eye, SciSports, Signality, Respovision, GradientSports and OptaVision. unravelsports turns tracking data into graph representations, pressing intensity and formation detection. Takes the corpus to 1,809 chunks across 23 providers.

Validation for string vocabularies in review

The checks that hold docs to a real package assume a vocabulary is made of identifiers — importable symbols, or SCREAMING_SNAKE enum members. Plenty of libraries use plain lowercase strings instead, and those docs were passing without being checked at all. Truth files now record what every Literal parameter accepts, so a documented coordinate system that does not exist fails the build.

Crawler fixes in review

Three, all found the hard way. Directory-style doc links were never followed, so mkdocs sites yielded their index page and stopped. HTML tables were being flattened one cell per line, which quietly destroyed every table in every crawled provider. And heading permalink anchors were being carried into chunk titles, so a URL was the first thing an agent read in a search result.

campos

IN BETA!

React components for football, typed and agent-native

23 React chart components, 10 data-provider adapters, one canonical coordinate frame, zero-config defaults. Typed from schema to render. Every chart is agent-discoverable via a static JSON registry at campos.withqwerty.com/r/ — nutmeg proposes and composes campos charts automatically when it detects a React project. Docs and live demos at campos.withqwerty.com.

chart components

data adapters

2333

tests

npm packages

Drives from nutmeg, automatically

When nutmeg detects a React project (package.json contains react or @withqwerty/campos-react), it loads a shared bridge doc, fetches the registry above, and proposes campos charts with real capability data per provider. Chart output, adapter selection, and recipe suggestion all route through the live catalogue — no hallucinated props, no training-data drift. See nutmeg on GitHub for the bridge reference.

nutmeg

Get started

football-docs

campos

Start building with football data