io.github.chrischall/myersparkathleticzone-mcp

myersparkathleticzone-mcp

MCP server for Myers Park High School (Mustangs) athletics — schedules, teams, rosters, coaches, news and game broadcast links.

No credentials. Everything this server reads is public. There is no login, no API key, no browser extension and nothing to configure before it works.

Developed and maintained by AI. Use at your own discretion.

Install

npx myersparkathleticzone-mcp

Or in an MCP host's config:

{
  "mcpServers": {
    "myersparkathleticzone": {
      "command": "npx",
      "args": ["-y", "myersparkathleticzone-mcp"]
    }
  }
}

Tools

Every tool is read-only; the server performs no writes.

Team-scoped tools need a numeric team id and a sport slug, and team ids differ per school year. Resolve them with mpaz_resolve_team rather than guessing — a stale id makes the site serve a Team … Not Found page (with HTTP 200), which the client detects and reports.

Configuration

Both optional:

The site is one tenant of the PlayOn Sports / SportsEngine "Athletic Zone" platform, so setting both points the server at another school on it. Verified against Ballantyne Ridge (https://www.ballantyneridgeathleticzone.com, school id 21785) as well as Myers Park.

Hosted deployments are Myers Park only. The hosted (mint / claude.ai connector) build runs behind a static egress allow list that permits just www.myersparkathleticzone.com, so setting MPAZ_SITE_URL to another school there makes every call fail with "Could not reach". Pointing at another school works only with a local stdio install (npm / .mcpb).

Known limits

These are properties of the upstream site, not bugs, and the tools say so in their output rather than guessing:

• A missing score is unknown, never zero. Each side's score is stored independently upstream, so half-entered games (3-null, null-4) are common. result is derived only when both sides are present.
• Myers Park records few scores. Every completed game of theirs observed so far has null scores, so a season may return fixtures with no results at all — that is the school's data entry, not a fault.
• Past-season coverage is partial and calendar-dependent. Myers Park's /schedule is empty for past years even though the per-team pages still serve them, so mpaz_list_teams falls back to the sport pages' cross-year team selector. That selector is keyed by team, not by sport slug, so it can only be reached through a team that is in season now — meaning off-season sports are unreachable for past years, and which ones are reachable changes through the year. The result carries a coverage note listing what was searched; an empty list there is not evidence the school fielded no teams.
• Most teams publish coaches but not players, so an empty roster is normal.
• The all-school schedule is a window of upcoming events, not a full season — use mpaz_get_team_schedule for that.
• Broadcast links are NFHS Network links, not hosted clips — durations are always 0.

How it works

The site is a fully server-rendered Next.js app whose own /api/* routes are server-only and reject outside requests. The data ships inside the page's React Server Components flight payload, so this server requests pages with the RSC: 1 header and harvests the entities embedded in them.

docs/ATHLETIC-ZONE-API.md pins the captured request/response shapes, the routes, the failure modes, and what was ruled out.

Development

npm install
npm run build
npm test

License

MIT