I gave my launch directory an mcp server so agents can look up products without scraping

I run ScrollLaunch, a weekly launch board for indie products. You list a product, it goes on that week's leaderboard, and it keeps its own page once the week's over. Agents and ai search mostly just need to read a page without it being a mess. So I tried to make that easy. The main thing is a public mcp server, read only, no API key needed. It has five tools. Search products, get the weekly leaderboard, get one product, get the startup directories ranked by DR, get launch stats. Point your agent at it and it can pull this week's launches or look up one tool, no HTML involved. Then there's the plain stuff. An llms.txt that's basically a short index, and an llms-full.txt with the full content of every product and blog post. Also a small json snapshot at /api/ai with this week's leaderboard and whatever went live recently. Every page has a markdown version too, you add .md to the url or send an Accept: text/markdown header. There's a REST API under /v1 with search, leaderboard, product detail and the directories list, plus an OpenAPI spec. Errors come back as problem details json so they're easy to parse. And a CLI, npx scrolllaunch, if you just want to poke at it from the terminal. Like the usual .well-known files for discovery too, an api catalog, an MCP server card, an agent skills index. It's rate limited so don't hammer it. And please don't scrape the site, that's why the other routes are there. Still figuring out what to add. If you're building agents that look up tools or products, what would you want these to return that they don't right now?