Markdown for Agents
Public pages on cvdportal.com serve markdown to any client that asks for it. This is content negotiation, so the URL never changes. Browsers keep getting HTML and agents get a clean document with no navigation, scripts, or layout markup around it.
Requesting markdown
Send Accept: text/markdown on a normal GET:
curl -H "Accept: text/markdown" https://cvdportal.com/faq/what-is-the-cyber-resilience-act
The response comes back as text/markdown; charset=utf-8 with a Link header carrying the canonical HTML URL and Vary: Accept so caches keep the two representations apart.
No API key is needed. These are public pages, and this is the same content the HTML renders.
Which pages support it
| Path | Content |
|---|---|
/ | Site index with a product summary and every public page, described |
/faq/<slug> | CRA FAQ answers with their sources |
/cra/<slug> | CRA article explainers |
/cra-guidance/examples/<slug> | Commission guidance worked examples |
/guides/<slug> | Industry guides |
/compliance/<slug> | Compliance checklists |
/for/<slug> | Role guides |
/eu/<slug> | EU country guides |
/glossary/<slug> | Glossary terms |
/tools/<slug> | Free tools |
/templates/<slug> | Policy templates |
/blog/<slug>, /news/<slug> | Posts and announcements |
Pages outside this list, such as /pricing and /features, return HTML whatever you send. Their content lives in application code rather than a content file, so there is no markdown form of them that would be more faithful than the page itself.
A slug that does not exist returns 404, exactly as the HTML page does.
Bulk alternatives
For ingesting the whole corpus in one request rather than page by page:
/llms.txtis the curated index, one line per page with a description./llms-full.txtinlines the full text of the CRA FAQ, the Commission guidance examples, and every article explainer.
Crawl and content-use policy
robots.txt allows every public page to 14 named AI crawlers and answer engines, and declares Content Signals for each:
Content-Signal: search=yes, ai-input=yes, ai-train=yes
All three are yes. Public CRA reference content exists to be found, quoted, and cited, including in AI answers. Dashboard, settings, billing, and submission routes stay disallowed for every crawler.
If you run a crawler and want a citation link, use the canonical URL from the Link header.