wc26.ai — 18 tools, zero API keys, works with Claude, ChatGPT, Cursor, and Telegram. Try it live.
Ask your AI about the 2026 World Cup schedule. You’ll get hedging, outdated info, or a web search that pulls random articles. That’s not useful.
So I built wc26-mcp — an MCP server that ships with everything bundled. 104 matches. 48 teams. 16 venues. City guides. Fan zones. Visa info. Head-to-head records going back decades.
One line to install: npx wc26-mcp
Why bundled data
The interesting decision was shipping all the data inside the npm package instead of hitting an external API. No API keys, no rate limits, no service to keep running. Install it once and it works offline.
World Cup data is mostly static. The schedule doesn’t change. Teams don’t change (much). City guides and visa requirements are stable. The only moving pieces are the 6 remaining playoff spots (March 26 & 31) and injury updates. For those, I push npm updates.
This pattern — curated static data compiled at build time — nobody’s really doing this with MCP. Most servers are API wrappers. But for bounded, high-quality datasets, bundling is simpler and more reliable.
The briefing tool
My favorite part. Say “brief me on the World Cup” with zero parameters. The server detects the current tournament phase and surfaces what matters today.
Right now: 120 days out, 6 playoff spots pending, here are the groups to watch. During the tournament it’ll shift to today’s matches, yesterday’s results, standings implications. The AI doesn’t need to know what to ask. It just calls what_to_know_now and gets a response that knows what day it is.
This is what MCP tools should feel like. Not “here’s an endpoint, figure out the parameters.” More like “tell me what I need to know” and the tool does the thinking.
Distribution beyond MCP
Not everyone runs MCP clients. So the same data powers three other surfaces:
- A ChatGPT GPT in the GPT Store — same 18 tools, no setup
- A Telegram bot with slash commands —
/team brazil,/city miami,/history argentina france - A REST API on Vercel for anyone who wants to build on top
Same bundled data underneath all of them. The MCP server, the GPT, the bot, and the API are all just different wrappers around the same data.
What I’m watching
The playoffs in March will be the first real content update cycle. Six teams qualify, I update the data, push to npm, and every installation gets current automatically on next install.
After that, trip planning. 16 cities across 3 countries is a genuine logistics problem, and AI is actually pretty good at figuring out logistics when you give it structured data to work with.