Public knowledge deserves to be used.

Folks. Sweden has two publicly funded organisations producing some of the country’s best insights on AI: RISE and AI Sweden. Reports, surveys, lessons from projects. Gold, actually. The problem is that most of it is hidden in PDFs and webpages built to be browsed, not used. Reasonably functional for people, but entirely unsuitable as a contribution to the emerging ecosystem orchestrated by agents.

So my AI team built what should have existed from the start: A completely open MCP server that makes both organisations’ public AI material searchable and directly accessible to any modern AI assistant. Claude, ChatGPT, Lovable, you name it. Paste a URL into your assistant, done. No login, no API key, no cost (to you). And if you’re a fellow nerd living in Codex, Claude Code, Cursor, VS Code or GitHub Copilot, you already know how useful this can be.

There are currently 2,675 documents (updated from 1,500, thanks to everyone who pointed that out) in Swedish and English. Webpages have been scraped and PDF contents extracted, and every answer links back to its source. Even if this is the first “MCP server” you’ve connected to your AI life, it takes a minute or less. Ridiculously simple, really. Detailed instructions are at the link in the comment below.

To my friends at RISE and AI Sweden: this was built with love. Your material deserves to be used. Feel free to take over the whole thing. It’s on me.

And to everyone sitting on a knowledge bank: this took a day to build and costs practically nothing to run. The question isn’t whether your knowledge should become readable by agents. It’s who gets there first: you, or someone like me 🤠

The image shows what triggered this build. A piece of a report (“State of AI by RISE”) that was only available as a PDF, while claiming to practise what they preach... You’ll find the link to all the instructions in the comments.

[ Update 2: the worst bug is fixed; everything should be back in place. I’m wondering whether to pull in EVERYTHING from RISE, though, not just material tagged as AI-related. What say you? ]