Footics MCP
Enables AI assistants to interact with the Footics World Cup 2026 prediction game, reading matches, standings, predictions, and optionally submitting predictions.
README
Footics MCP
A remote MCP server for Footics,
a World Cup 2026 prediction game. A user connects their AI assistant (Claude,
ChatGPT, Cursor…) to https://mcp.footics.app/mcp, signs in with their Footics
account, and the assistant can read their matches, standings and predictions —
and, optionally, submit them. Inference runs on the user's side; Footics only
pays for hosting.
Separate Vercel project from the main site, same Supabase database. Built with
Next.js 16 (App Router) and mcp-handler.
Architecture
AI assistant (Claude / ChatGPT) ──MCP/HTTP──► app/[transport]/route.ts (/mcp)
│ withMcpAuth → lib/auth (verify Supabase JWT)
│ lib/tools → lib/queries (reads)
│ → lib/predictions (writes, gated)
▼
Supabase Postgres (shared with the site)
lib/queries.ts— lean reads that return raw JSON for the model, not the UI view-models.lib/predictions.ts— writes with the same guardrails as the main app. Off by default (MCP_ENABLE_WRITES=false).db/schema.ts,db/index.ts,lib/jokers.ts— mirror the main app's schema and rules so column contracts never drift.
Tools
whoami · list_matches · get_match · get_my_standing · get_my_predictions ·
get_leaderboard · list_my_groups · get_joker_status · search ·
submit_prediction (gated)
Configuration
| Variable | Default | Effect |
|---|---|---|
MCP_ENABLED |
true |
false → /mcp returns 503 with no Supabase/DB call (kill switch) |
MCP_REQUIRE_AUTH |
true |
require a valid bearer token (401 otherwise) |
MCP_ENABLE_WRITES |
false |
allow submit_prediction |
MCP_TOKEN_VERIFY |
getuser |
getuser (robust) or jwks (local, asymmetric) |
MCP_RATE_LIMIT_PER_MIN |
30 |
tool calls per minute per user (0 = off) |
See .env.example for the full list, including the Supabase
connection variables.
Development
pnpm install
cp .env.example .env.local # fill in — point at a staging project, not production
pnpm dev # http://localhost:3000/mcp
pnpm token # mint a test access token + ready-to-paste MCP Inspector commands
pnpm token prints an MCP Inspector command: run tools/list, then whoami,
then any tool. Without an Authorization header the endpoint returns 401.
Security
Auth is required and writes are off by default. See SECURITY.md for the model and how to report a vulnerability.
License
MIT © 2026 Tom Cardoen
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。