Using this with an AI agent
Markdown, llms.txt, and what to hand a model so it answers correctly.
Every page on the main site is published twice: as HTML, and as markdown at the same path with .md appended. The markdown carries the law, the streams, the prices and the full question set, so a model that reads one page has enough to answer without guessing.
Give a model the pricing page
curl -s https://eprfrancerep.com/pricing.md
Read from https://eprfrancerep.com/pricing.md and answer using only what it says.
Index of everything
https://eprfrancerep.com/llms.txt lists every page on the site.
Machine-readable reference
https://eprfrancerep.com/api/v1/index.json is the discovery document: it points at the four reference files and describes the client API.
The reference files are regenerated whenever the site builds, so they cannot drift from what the pages say.