I built a small MCP server where your AI can read control and concepts DOCS. Feel free to use it if you find it helpfull.
There is nothing fancy there. Docs are downloaded as you work with them and then cached.
This looks great, I will definitely try it. We used Context7 for this previously, but this looks easier.
Actually, we started doing something similar - we create a skill that helps AI to generate DotHTML markup without hallucinating components and their properties, and we gave it “compressed” documentation to prevent the most common mistakes - just a few markdown files with most useful tips and tricks on how to write components, how to do validation, and so on.
We need to carry out some benchmarks to see how well it performs.